Add tests for replica deletetion.

Try do delete corresponding replica tuple from master cluster space and
try to join replica to cluster. Master and replica should not crash,
replication should be stopped.