kolla_action=stop should play in reverse order

[Ettore Simone]

The current implementation of `Kolla-ansible stop` command is executed in the same order as `kolla-ansible deploy`.

This implies that databases, memcache, and message queue services are stopped before the containers are able to gracefully shut down, inducing a much longer time to shutdown the complete stack and often signaling the containers due to timeouts (in stuck because are not able to communicate with shared services).

I'm wondering if is the case to implement a version of "kolla-ansible stop" that shuts down containers in reverse order.