Skip update of relay-log.info in GTID mode

In GTID mode, the current slave pos is stored crash-safe in
mysql.gtid_slave_pos. There is no need to update the relay-log.info file after
every event group, which is not crash-safe anyway and costs performance
(write(2) syscall for every transaction).

Especially with parallel replication, the update of relay-log.info could
start to hurt.

It could perhaps be an option, in case some existing monitoring solutions want
to read the file from outside the server rather than use SHOW SLAVE STATUS.