Benchmark new strategy for non-co-located and co-located nodes for cassandra streaming

These are all outdated tickets, closed by me last for a reason. The new tickets are all in the new repo as noted on the reviewer README.

the work is simply in timing output logged when running tests for 5k write/read sequences for now. I had to get something started to insure the implementation was doable WRT a basic SLA.

1. This test shows for 5000 writes/reads, for each, write, the streamed read is received in 0 - 2 ms: https://github.com/riptano/cassandra-streaming/blob/wip-cassandra-input-dstream-v2/cassandra-streaming-tests/src/it/scala/com/datastax/spark/connector/streaming/StreamingTriggerIntegrationSpec.scala#L93-L109

2.A. Logging shows that new data is received here:
https://github.com/riptano/cassandra-streaming/blob/wip-cassandra-input-dstream-v2/cassandra-streaming-publisher/src/main/scala/com/datastax/spark/connector/streaming/StreamingTrigger.scala#L70
2.B. Returned to Cassandra here
https://github.com/riptano/cassandra-streaming/blob/wip-cassandra-input-dstream-v2/cassandra-streaming-publisher/src/main/scala/com/datastax/spark/connector/streaming/StreamingTrigger.scala#L77
2.C. in 0 ms

What is to be reviewed here? There is no link to a PR or anything...

Testing triggers with the connector and EmbeddedCassandra is proving a bit tricky, doing some custom / new setup.