avoid calling getThroughSeqnos and getHighSeqnoAndVBUuidFromTarget more than once

In the fix for MB-21812, we call getHighSeqnoAndVBUuidFromTarget() and getThroughSeqnos() instead of commitForCheckpoint() to avoid disk commit on target.  these methods are called only once when we do periodical checkpointing, which is expected. However, they can be called more than once when we do one time checkpointing before pausing replication. This needs to be fixed.