Parse, resolve, write and compare URLs the way browsers do, by the URL
Standard.

A URL is a plain map of the parts it has:

- :scheme, in lower case, e.g. "https"
- :username and :password, percent-encoded, when they aren't empty
- :host, as text, e.g. "example.com", "127.0.0.1" or "[::1]" (it
  can be empty, as in file:///etc/hosts)
- :port, a number, unless it's the default of the scheme
- :path, a vector of segments, or a string for an opaque path like that
  of mailto:me@example.com
- :query and :fragment, without their ? and # (both can be empty)

Every function that takes a URL also takes its text (e.g. a string or a
java.net.URI) and parses it as an absolute URL first. For text that
doesn't parse, it returns nil.

The functions named after the attributes of JavaScript's URL class (e.g.
hostname and pathname) return their values. To set them the way that
class's setters do, use with and update.

The hosts of URLs by the URL Standard, section 3: domains, IPv4 and
IPv6 addresses, opaque hosts and the empty host.

A host is the text that parse returns for it, e.g. "127.0.0.1" or
"[::1]". The functions that end in -report also return the validation
errors as keywords. Pass them a vector to add the errors to, or nil to
skip them.

Domain names in Unicode and in ASCII, by Unicode IDNA Compatibility
Processing (UTS #46) of Unicode 18.0.0, revision 36.

The options are those of UTS #46, and default-options holds the ones
that the URL Standard uses for the domain of a URL. NFC comes from the
platform. A domain is an error rather than a guess when it has a new
code point that NFC doesn't leave alone and the platform doesn't know.

Query strings and form bodies in application/x-www-form-urlencoded, by
the URL Standard, section 5, with the methods of URLSearchParams from
section 6.2.

The pairs are a vector of name-value vectors in order, in which a name
can come more than once. Each function also takes a map or any other seq
of pairs, e.g. {:q "clojure" :page 2}. A keyword is read as its name
(:q as q) and any other value as its text.

Percent-encoding and percent-decoding, by the percent-encode sets of
the URL Standard, section 1.3.

Each set is a predicate of a code point, so a set of your own is a
function too, e.g. (some-fn percent/path-set #{0x3B}). Every set of the
standard holds the code points above U+007E.

TODO: the encodings other than UTF-8 that HTML passes to the parser for
the query of a document in a legacy encoding.

Punycode by RFC 3492, which writes a Unicode label of a domain name in
ASCII, e.g. the xn--fa-hia of xn--fa-hia.de for faß.de.

Like most implementations, it limits its numbers to 31 bits.