Offline replicators should be allowed to have errors

There is a discussion on how to fix CBL-2354 (a linked ticket) and I believe the correct fix is to not check the error unless the replicator is stopped.  If a replicator went offline, it had an error by definition (or else it wouldn't be offline).  It appears, however, that when the test framework for mobile waits for a replicator to stop, any error at any time will cause a failure regardless of status.  This causes incorrect conclusions about the state of the library (namely this time around that the library "is not retrying.")