Additional MTR tests for backup locks

Taken from https://code.launchpad.net/~akopytov/percona-server/backup-locks/+merge/206724/comments/487927. This is also Jira PQA-70.

    - Need a test that checks for a LTFB/LBFB failure if issued by a
      user without RELOAD privilege.

    - Need a test for crash-safe slave: slave server binlog on, log
      slave updates off, --master-info-repository=TABLE,
      --relay-log-info-repository=TABLE, LOCK BINLOG on the slave,
      verify that updates from the master are not blocked.

    - I was not able to find a test where a single connection would
      take either backup lock, and attempt to issue FTWRL.

    - Some cases which would be interesting to test, but IMHO less
      value than above:

    - EVENTs. Both that CREATE EVENT is blocked on mysql.event or
      its corresponding binlogging, and that firing EVENT obeys the
      locks.

    - XA transactions and more statements that write to mysql schema:
      CREATE USER, GRANT, etc.

KAlexey:
   - triggers
   - LOCK TABLES table_list READ/WRITE + LOCK TABLES FOR BACKUP in either the same, or another connection
   - FLUSH TABLES table_list WITH READ LOCK + LOCK TABLES FOR BACKUP in either the same, or another connection
   - multiple concurrent LTFB/LBFB
   - LTFB/LTFB + BINLOG statements
   - stored routines + LTFB
   - prepared statements + LTFB
   - test LTFB (session 1) + DDL (session 2) + SELECT (session 3)
   - test LTFB + INSERT INTO myisam when table is already in TC
   - test LTFB + INSERT INTO myisam + FLUSH TABLES + INSERT again
   - all combinations of online/inplace ALTER TABLE + LTFB
   - multi-table updates + LTFB
   - CREATE TABLE IF NOT EXISTS under LTFB
   - implicit updates to mysql.* tables (CREATE USER, GRANT, etc.) + LTFB/LBFB
   - partitioned tables with both "online" and "offline" storage engines as partitions + LTFB
   - FLUSH LOGS + LTFB/LBFB

Completed in https://code.launchpad.net/~akopytov/percona-server/backup_locks_bugs/+merge/235794:

    - I am not too familiar with MTS slave, but I see MTS
      coordination calls next to binlog protection in
      Rotate_log_event::do_update_pos. Is it worthwhile to add a
      simple MTS testcase?