Problem:
--------

The NDB index stat tables are missing from the MySQL Server after
an initial system restart. The binlog thread of the first MySQL
Server connecting to the Cluster post an initial restart pokes
the index stat thread to inform it about the initial restart.
The index stat thread then recycles its Ndb object and creates the
index stat tables in NDB. The binlog thread signals the index stat
thread after the synchronization phase which results in the tables
not being synced to the DD of the first MySQL Server connecting to
the Cluster.

If there are multiple MySQL Servers connecting to the Cluster,
there's a race condition during subsequent MySQL Server connections
where the index stat tables could be synchronized or not depending
on when the index stat thread creates the tables in NDB. If the
creation occurs in the middle of the sync during binlog setup of a
Server, it opens the door to unpredictable behaviour and errors in
the sync of these tables.

Fix:
----

The ndb_index_stat_head and ndb_index_stat_sample tables aren't
exposed to the MySQL Server. This makes them exempt from all types
of metadata synchronization and table discovery. The presence of
these tables in the MySQL Server offers little value to users since
their contents are incomprehensible without some kind of parsing.
The tables remain visible and accessible via NDB tools such as
ndb_show_tables, ndb_desc, and ndb_select_all.

Change-Id: I30bdce6451cc1c99b6f7dc982fc634ddcce61404