futures that clear conveyed bindings after execution.

Fixes memory leak described in https://clojure.atlassian.net/browse/CLJ-2619

An implementation of clojure.core/every-pred with a simple definitional equivalence.

Variant of `clojure.core/areduce` that supports naming the array.

Variants of clojure.core functions (and internal helpers)
that release the head of seqs earlier, enabling lower peak memory
usage in some cases.

For example, many higher-order functions in this namespace release strong
references to arguments before calling their function arguments.
Realizing the following example would be prone to failure with an OutOfMemoryError using
clojure.core/map because map retains a strong reference to takes-75-percent-of-heap
when calling its function argument, which prevents memory from being reclaimed to create
another value of that size.

(map (fn [takes-75-percent-of-heap]
       (if (< (rand) 0.5)
         takes-75-percent-of-heap
         (generate-another-value-taking-75-percent-of-heap)))
     [takes-75-percent-of-heap])

In contrast, using head-releasing/map, the garbage collector can reclaim takes-75-percent-of-heap
while calculating (generate-another-value-taking-75-percent-of-heap) because
head-releasing/map does not strongly reference takes-75-percent-of-heap at that point.

The basic implementation trick to achieving this is to call (rest s) on the seq currently
being processed _before_ calling (f (first f)), so the strong reference to (first f)
is transferred from the higher-order function to f during the call to f.

There are potential caveats to this approach: https://clojure.org/reference/lazy#_extension_iseqs
If the underlying ISeq implementation defines rest in terms of next as in the article, then the
functions in this namespace will force two seq elements into memory simultaneously.
For example, the call below will throw an OutOfMemoryError before the fn is called because both
elements of the seq will be realized.

(map (fn [takes-75-percent-of-heap] nil)
     (lazy-seq-where-rest-calls-next
       (cons (->takes-75-percent-of-heap)
             (lazy-seq [(->takes-75-percent-of-heap)]))))

Infinite lazy seqs such as cycle or repeat always hold strong references to all their elements, so
the functions in this namespace will have no affect in the memory usage of processing these seqs.

Another caveat is that the functions here are perhaps more useful as
validation of the leaky-seq-detection framework and for pedalogical purposes about sequences
than as significant bump in real-world expressivity.
See the tests for how these implementations were verified:
  io.github.frenchy64.fully-satisfies.leaky-seq-detection-test

The author of this namespace can only speculate why the original functions were written this way.
Perhaps the idea of a fn releasing a strong reference to one of its arguments was too rare to risk
the caveats in using the default implementation of ISeq. Chunked seqs are also so prevalent
and have much higher memory requirements that the optimizations in this namespace might
have been deemed insignificant. For example, map processing a chunk of 32 must have enough memory
to hold both 32 elements of the input collection and 32 elements of the output collection simultaneously.
Chunked seqs also make programs such as the above unidiomatic since they are so difficult to get right
(see the hoops math.combinatorics jumps through to reduce the 32+32 elements of the previous example to
32+1).

At the very least, this work helps crystallize the differences between rest and next---or
more precisely, their similarities: while rest does not realize the next element, nor tell us whether
a seq has more elements, it is still an eager operation whose result, like next, releases a strong
reference to the first element of the seq.

int n = 0;
while(iter.hasNext() && n < CHUNK_SIZE)

int n = 0;
while(n < CHUNK_SIZE && iter.hasNext())

Variants of clojure.core functions that are slightly lazier when
processing and/or returning lazy seqs.

Some common (anti)patterns were to blame for the less-lazy versions of
these functions. The main insight was that if you are using a counter to
bound how many elements of a seq you are walking, you can be lazier by
testing the counter _before_ testing for more elements. For example, 

```java
int n = 0;
while(iter.hasNext() && n < CHUNK_SIZE)
```

will force n+1 elements and

```java
int n = 0;
while(n < CHUNK_SIZE && iter.hasNext())
```

will force n elements. In this case, realizing the extra element has no utility because
the chunk only fits CHUNK_SIZE elements.

Another problematic pattern was using seq-based implementations for sequence functions.
Seqs must be non-empty, but a sequence can be empty, so implementations can be lazier.

A framework to detect memory leaks caused by holding onto the head of sequences.

The java.lang.ref.Cleaner class (JDK9+) provides hooks into garbage
collection. You can register a function that is called when a value
becomes phantom reachable, indicating is it a candidate for garbage
collection.

The JVM is very likely to perform garbage collection
right before throwing an OutOfMemoryError. Part of garbage
collection is calculating whether references are reachable.
We use this insight to force garbage collection (and hence, cleaners)
to run, by inducing an OutOfMemoryError in try-forcing-cleaners!.
Note that an OutOfMemoryError can leave the JVM in a bad state, so
this strategy is best isolated away from other tests.

Tying these ideas together are reference-counting seqs and the is-strong
testing macro. ref-counting-lazy-seq returns a lazy seq
and an atom of all elements of the seq currently with strong references.
This seq can now be passed to a sequence-processing function you would
like to test for memory leaks.

is-strong then takes a set of seq indexes expected to have strong references
and checks them against the atom tracking strong references. It will continue
to induce OutOfMemoryError's until the expected references are found, or eventually
fail via an is test assertion.

Note that it's best to build up the set of expected strong references rather than
whittle it down. From a usability standpoint, starting with (is-strong #{} live)
will likely print the actual set of strongly referenced indexes, from which you can
source the expected strong references (from the result itself or a subset).
But false-positives are possible if an overly broad superset of the actual strong references are provided,
since is-strong may short-circuit its search earlier than the cleaners can run since.
Such results can lead to false conclusions about how lazy a function is, especially
if you have a pre-conceived notion of the results! The JVM property
  io.github.frenchy64.fully-satisfies.leaky-seq-detection.is-strong.false-positive-detection=true
will more aggressively search for such false-positives, which may be suitable for a cron CI job.

Here's an example of asserting that a program adds or subtracts strong references to elements
of a lazy seq at particular points.

(deftest example-cleaners-test
  (let [{:keys [strong lseq]} (ref-counting-lazy-seq
                                {:n 10}) ;; seq of fresh Object's, length 10
        ;; lseq=(...)
        _ (is-strong #{} strong) ;; no elements currently in memory
        lseq (seq lseq)
        ;; lseq=(0 ...)
        _ (is-strong #{0} strong) ;; just the first element in memory
        _ (nnext lseq)
        ;; lseq=(0 1 2...)
        _ (is-strong #{0 1 2} strong) ;; the first 3 elements in memory
        lseq (next lseq)
        ;; lseq=(1 2 ...)
        _ (is-strong #{1 2} strong) ;; the second 2 elements in memory
        lseq (rest lseq)
        ;; lseq=(2 ...)
        _ (is-strong #{2} strong) ;; the third element in memory
        lseq (rest lseq)
        ;; lseq=(...)
        _ (is-strong #{} strong) ;; no elements in memory
        ;; lseq=(3 ...)
        lseq (seq lseq)
        _ (is-strong #{3} strong) ;; fourth element in memory
        _ (class (first lseq)) ;; add a strong reference to lseq so previous line succeeds
        ;; lseq=nil
        _ (is-strong #{} strong) ;; lseq is entirely garbage collected
        ]))

See io.github.frenchy64.fully-satisfies.leaky-seq-detection-test for real-world
examples of finding memory leaks in Clojure functions, and then verifying fixes for them.

Linear-time sequence functions

Implementations of clojure.core macros that don't leak implementation details.

Implementations of clojure.core.async macros that don't leak implementation details.

Implementations of clojure.core.logic.pldb macros that don't leak implementation details.

Implementations of clojure.core.match.debug macros that don't leak implementation details.

Implementations of clojure.java.jmx macros that don't leak implementation details.

Implementations of clojure.java.test macros that don't leak implementation details.

Implementations of clojure.pprint macros that don't leak implementation details.

Implementations of clojure.test macros that don't leak implementation details.

Implementations of clojure.test.check.generators macros that don't leak implementation details.

Implementations of clojure.test.check.properties macros that don't leak implementation details.

Implementations of clojure.test.tap macros that don't leak implementation details.

Implementations of clojure.tools.trace macros that don't leak implementation details.

Provides a spec for clojure.core/reify via ::reify-args.

To register a spec for reify, call (register-reify-spec).

An alternative to `clojure.core/run!` that does not short-circuit on reduced.

Variants of clojure.core functions that improve thread-safety and general robustness
when passed mutating collections.

We agree that 'Robust programs should not mutate arrays or Iterables that have seqs on them.'
https://clojure.org/reference/sequences
Eductions inhabit a middle ground. They are designed to be walked from first to last like seqs,
but each element is recomputed when 

For example, clojure.core/split-at could disagree on the take/drop parts of
the collection if the coll is mutated between realizing the splits.
Here, any one call to the eduction alternates between [0 1 2 3 4 5 6 7 8 9] and
[9 8 7 6 5 4 3 2 1 0]. However, split-at incorrectly splits the eduction as
[[0 1 2 3 4] [4 3 2 1 0]], and safer/split-at correctly splits it as
[[0 1 2 3 4] [5 6 7 8 9]].

(deftest split-at-mutation-test
  (let [up-down (atom true)
        ed (eduction (map (fn [i]
                            (when (zero? i)
                              (swap! up-down not))
                            (if @up-down
                              (- 9 i)
                              i)))
                     (range 10))]
    (is (= [[0 1 2 3 4] [4 3 2 1 0]] (split-at 5 ed)))
    (is (= [[0 1 2 3 4] [5 6 7 8 9]] (safer/split-at 5 ed)))))

See io.github.frenchy64.fully-satisfies.safer-test for more details.

The basic trick here is strategically calling seq earlier on the collection argument.

An implementation of clojure.core/some-fn with a simple definitional equivalence.

Drop-in replacements for `clojure.test/{deftest,testing}` that (when used
together) enhances uncaught exception error messages with the (most likely)
testing context it was thrown from.

Example:

  (deftest my-test
    (testing "foo"
      (doseq [v [1 2 3]]
        (testing v
          (assert (= 1 v))))))

With clojure.test/{deftest,testing} 1.10.3 (notice `foo 2` is not mentioned):

  user=> (test-var #'my-test)
  ;ERROR in (my-test)
  ;Uncaught exception, not in assertion.
  ;expected: nil
  ;actual: java.lang.AssertionError: Assert failed: false
  ;...
    
With {deftest,testing} in this namespace (notice `foo 2` is mentioned):

  user=> (test-var #'my-test)
  ;ERROR in (my-test)
  ;Uncaught exception, possibly thrown in testing context: foo 2
  ;expected: nil
  ;actual: java.lang.AssertionError: Assert failed: false
  ;...

Variants of clojure.core functions that are generalized
to work uniformly for all values.

In all cases, a namespaced keyword was used as a special value
that, if provided to the function, would break its promised semantics.
The fixes involved replacing these special values with globally unique
ones that are inaccessible to normal users (or more practically, unlikely
to be generated with a generator like gen/any, or one that sources its
values from the keyword interning table).

These corner-cases are demonstrated in io.github.frenchy64.fully-satisfies.uniform-test.

An effective generator that could find such defects reliably could generate
keywords that occur in the source code of the functions reachable from the
generative property (using static analysis, but perhaps this is also retrievable
dynamically from the bytecode). On the other hand, such invasive analyses could
yield false-negatives by providing values normally inaccessible to the user.