This worklog implements the monitoring of replication lags using the
replication P_S tables.
The new monitoring information added consists of the GTID, the original
and the immediate commit timestamps of the transaction that is currently
being processed (plus the corresponding start processing timestamp) and
of the last to be processed (plus the corresponding start and end
processing timestamps) in each of these three tables:
- performance_schema.replication_connection status
- performance_schema.replication_applier_status_by_coordinator
- performance_schema.replication_applier_status_by_worker

All timestamp fields in these three tables have microseconds precision,
including last_error_timestamp and last_heartbeat_timestamp.
The field 'last_seen_transaction' in table
performance_schema.replication_applier_status_by_worker was replaced by
'applying_transaction' and 'last_applied_transaction'.