Support retry control

[fruch]

We are looking at using latte in some of our tests, and those test are chaos originated, and can take a node down, or grow/shrink/upgrade the cluster while a tool like latte (or cassandra-stress) is working.

cause of that we'll need some why to control retry requests

• if in the script itself
• if in some command line options, similar to cassandra-stress -error retries=20