OCTF DB replication must check for A/P cluster

[Bogdan Dobrelya]

Related bug https://bugs.launchpad.net/fuel/+bug/1585506 contains a failures that I called the "Type 1", which is a table cannot be found on a node != the node it was created on.

It places the error to ostf log like:
Command 'mysql -h localhost -e "SELECT * FROM ost1601.ost474 WHERE data = "9026077276"" ', exit status: 1, Error:
ERROR 1146 (42S02) at line 1: Table 'ost1601.ost474' doesn't exist

The test case as is may be good only for A/A cluster, while Fuel sets up A/P. And it must be reworked as the following:
* Make sure the nodes you use for queries are Active and not a Standby or down HAProxy backends.
* Rework the select query to be executed on a newly elected Active backend, for example:
 - Create table on the active backend foo
 - Fail the foo and use new active backend bar (wait for HAProxy to finish fail-over). Note, if it's not doable as OSTF shall not be destructive, we have to just remove the replication test then, or make it for only A/A Galera, which is all healthy backends are Active.
 - Check if the table exists on the backend bar

* Test 1: Use tx_isolation = repeatable read and run all operations as DB transactions.
* Test 2: if the test 1 fails, re-try it with the session-level set http://galeracluster.com/documentation-webpages/mysqlwsrepoptions.html#wsrep-sync-wait
* Report failure if only both test 1 and test 2 has failed