Document learning and best practices on usage/costs/etc of DB services

[fperez]

Thanks to work by @abbyazari @consideRatio & team, we ended up with a great setup for being able to run persistent DB servers on our Hub. This is a different use case from the ephemeral, on-demand spinning of instances for interactive usage. But having systems like this is a very real need for many scientific workflows.

Our setup worked, but it might also have been a bit sub-optimal in terms of cost/etc. This is to track that we document how this can be done in JHub and what suggestions there may be for others - I'm sure many scientists will appreciate this.