Problem with MySQL Server holding the global read lock
while calling ha_reset_logs(), this prevents ndbcluster
from modifying data in its ndb_binlog_index table in RESET MASTER.

Fix by changing ndbcluster to use the
Binlog_transmit_observer::after_reset_master
hook in addition th ha_reset_logs(). Thus the code which
modifies data in the table is done after global read lock
has been released.

Improve comments and function description for
Ndb_binlog_thread::do_after_reset_master() which is now the
callback executed when the MySQL Server does a RESET MASTER.

Remove condition to clear ndb_binlog_index only if ndbcluster is writing
binlog, should be done unconditionally.

Improve ndbcluster_binlog_wait(), update function comment, reverse logic
to exit the function directly when not binlogging. Improve DBUG_PRINT's

NOTE! This patch depends on upstream fix to release the
global read lock before calling RUN_HOOK(after_reset_master)

Change-Id: I7de69235132e02d9e9785ac7c0ace280dba8c852