Clojure Redis client & message queue.

EXPERIMENTAL support for Redis Cluster atop Carmine.

Define an appropriate function for EVERY Redis command. This is done by
parsing the official Redis command reference (JSON) which includes up-to-date
docstrings, argument specs, etc. This awesome approach was adapted from
labs-redis-clojure.

Handles socket connection lifecycle. Pool is implemented with Apache Commons
pool. Originally adapted from redis-clojure.

Alpha - subject to change.
Distributed lock implementation for Carmine based on work by Ronen Narkis
and Josiah Carlson. Redis 2.6+.

Redis keys:
  * carmine:lock:<lock-name> -> ttl str, lock owner's UUID.

Ref. http://goo.gl/5UalQ for implementation details.

Carmine-backed Clojure message queue. All heavy lifting by Redis.
Message circle architecture used here is simple, reliable, and has
reasonable throughput but at best mediocre latency.

Redis keys:
  * carmine:mq:<qname>:messages     - hash, {mid mcontent}.
  * carmine:mq:<qname>:locks        - hash, {mid lock-expiry-time}.
  * carmine:mq:<qname>:backoffs     - hash, {mid backoff-expiry-time}.
  * carmine:mq:<qname>:nattempts    - hash, {mid attempt-count}.
  * carmine:mq:<qname>:mid-circle   - list, rotating list of mids.
  * carmine:mq:<qname>:done         - set, awaiting gc, requeue, etc.
  * carmine:mq:<qname>:requeue      - set, for `allow-requeue?` option.
  * carmine:mq:<qname>:eoq-backoff? - ttl flag, used for queue-wide
                                      (every-worker) polling backoff.
  * carmine:mq:<qname>:ndry-runs    - int, number of times worker(s) have
                                      burnt through queue w/o work to do.

Ref. http://antirez.com/post/250 for basic implementation details.

Facilities for actually communicating with Redis server using its
request/response protocol. Originally adapted from Accession.
Ref. http://redis.io/topics/protocol

Carmine-backed Ring session store. Adapted from clj-redis-session.

Semi-automatic datastore layer for Carmine. It's like the magix.
Use multiple Redis instances (recommended) or Redis databases for local key
namespacing.

Redis keys:
  * carmine:tundra:evictable -> set, keys for which `ensure-ks` fetch failure
                                should throw an error.

Secondary Redis server DataStore implementation for Tundra.

Simple disk-based DataStore implementation for Tundra.

Faraday (DynamoDB) DataStore implementation for Tundra.

Use AWS Data Pipeline to setup scheduled backups of DynamoDB table(s) to S3
(there is a template pipeline for exactly this purpose).

AWS S3 (clj-aws-s3) DataStore implementation for Tundra.