Possible fix for Cassandra / Kairos problems on staging?

[timcowlishaw]

Not for merging because untested (and based off the new compose file branch which isn't yet merged), but this might be a simple solution to #268 - give cassandra a healthcheck in the docker compose file which would allow it to reboot if it becomes unresponsive, and specify that kairos depends on it being healthy, so it cleanly starts and restarts in response.

[oscgonfer]

Testing in staging. The outcome is that kairos waits for cassandra-1 to be healthy and then starts. So far so good!