Containerized deployment

We arguably ought to be deploying shaman as a container. To start, we could convert just shaman, but eventually we may want to deploy postgres and rabbitmq as containers as well.

[paddles](https://github.com/ceph/paddles) might be a good starting reference point.

Start by reviewing how shaman is currently deployed and use paddles as the stated reference point. Define the initial containerized scope for shaman separately from the possible future postgres and rabbitmq containers; the work is done when that scope and its deployment process are documented and usable.

Written by the indexing model from the issue text.