Verify failover behavior with semi-sync wait point set to AFTER_SYNC

Test that in the following configuration, a failure of the primary server ends up with a failover to the most up-to-date server.

• in MaxScale, mariadbmon is configured with
  • cooperative_monitoring_locks=majority_of_all
  • auto_failover=true
  • auto_rejoin=true
  • enforce_simple_topology=true
  • enforce_writable_master=true
  • enforce_read_only_slaves=true
• in MariaDB
  • rpl_semi_sync_master_wait_point=AFTER_SYNC
  • --init-rpl-role=SLAVE
  • read-only=1

The test is to kill the primary server with something like pkill -11 mariadbd and then to check that failovers can be done without the old primary being unable to rejoin the cluster. During this testing, there should be intense write traffic through MaxScale and all successfully committed transactions (i.e. client gets the OK for the commit) should be durable. Users with SUPER or READ_ONLY ADMIN privileges must not be used as those bypass the protection that read_only=1 gives.