Come up with Reference Plugins for Benchmarking

@suyograo We spoke about this last week, we should have a strong benchmarking strategy based on actual plugins of known (and preferably very small) overhead.

I think one clean way to do this, would be optimizing the [TCP Input](https://github.com/logstash-plugins/logstash-input-tcp/) and [output](https://github.com/logstash-plugins/logstash-output-tcp/) plugins both.

Then we could simply run `nc -l ${port} > /dev/null` as a sink on the output side and blast the input side with some fixed data in a loop on a loopback `SocketChannel`.

Just to keep track of this exchange :) Not sure if you want me or anyone, in general, to put some time towards stronger benchmarks at this point.