initial draft implementation of thundering herd mitigation

[jsvd]

Alternate solution for #377

Currently the beats input will accept connections and pass the work of decoding the message to an executor that has an unbouded queue.

This makes it so that the number of batches in flight in Logstash's off heap memory will equal the number of beats (more if pipelining is enabled).

TODO:

• add metrics when limits are hit
• add trace logging
• describe solution
• document how to test
• consider making limits configurable