Add ability to auto-discover cluster

[jbeemster]

This would be based on name and age where rather than passing the jobflow ID you would pass the jobflow name and dataflow-runner would select the newest cluster available that is running.

Thoughts @alexanderdean @BenFradet ?

[alexanderdean]

I really like this:

1. It could allow us to avoid hardcoding jobflow IDs (which change as we replace clusters) into DAGs
2. It allows us to do blue-green switchovers

[chuwy]

Just another possible way, probably easier to implement: #35

[jbeemster]

That could work as well - could be an idea that you could also store a jobflow-id as a Consul KV pair to be checked also...