MTR hangs when bringing up a Galera cluster fails

A single galera/wsrep test stalls the whole MTR run until suite timeout expires. The --testcase-timeout never triggers because the hang happens before the testcase timer is started.

Root cause: When bringing up the cluster, start_servers blocks in wait_wsrep_ready until each non-bootstrap node reports wsrep_ready=ON. If a server hangs during startup and/or stops responding to wsrep_ready queries, the whole MTR worker hangs.

Suggested fix: Make the wsrep readiness wait time bound and stop waiting on hung client.