Comment 23 for bug 1208493

Merged the fix for wsrep_cluster_address update into wsrep-5.5, and added two more fixes for wsrep_provider changes:
* counting applier threads by wsrep_running_threads variable, which is accessed under LOCK_thread_count mutex
    There was a race condition, when appliers were counted by scanning threads, which did not hold LOCK_thread_count
* avoiding to put wsrep applier threads back to thread cache. Appliers' closing does not currently send correctly
    COND_thread_cache signal, and applier might hang when appending to thread cache

These fixes were pushed in revision: http://bazaar.launchpad.net/~codership/codership-mysql/wsrep-5.5/revision/3936
and later cleanups in following revisions #3937 and #3938