Added purge-binary-logs/:host/:port/:logFile API endpoint, supporting ?force=true

purge-binary-logs will now refuse purging, by default, if the host has replicas which have not yet applied events in the binary logs to-be-purged.

This is a strict sanity check. A more relaxed one would be to check if the replicas have copied the binary logs as relay logs. This is good enough on a normal flow, but since relay logs are purged on a change master to command, they can lose their ability to replicas. So orchestrator plays safe.