Log aggregation finish time should get logged for trouble shooting.

Now, RM track application log aggregation status in RMApp and the status change is triggered by NM heartbeat with log aggregation report. For each node's log aggregation status change from in-progress (NOT_START,RUNNING,RUNNING_WITH_FAILURE) to final status (SUCCEEDED,FAILED, TIMEOUT), it trigger an aggregation for log aggregation status: updateLogAggregationStatus(). The whole progress is log less and we cannot trace the log aggregation problem (delay of log aggregation, etc.) from RM (or NM) log. We should add more log here.