Graceful handling of shutdown and restart

[sharon-tickell]

This PR builds on #25, going beyond just a version update to perform some refactoring of the controller-logic.

It refactors the replica_ctrl.py script to:

• NOT cache and reuse the Mongo service instance, but instead re-retrieve it on each scheduled check. This means the mongo_controller service doesn't need to exit and restart if the mongo service it is controlling has restarted: it just handles that situation gracefully.
• Force-update the replicaset configuration when more than one replicaset member needs to be changed. (It would previously fail, exit and restart in this scenario)
• Gracefully exit the control-loop if a SIGINT or SIGTERM signal has been sent to the replicaset, indicating that the docker service is being shut down.