K8s: Investigate more robust way of handling snapshots

Created by: ggilmore

(This is a stub issue, I'll flesh it out later)

We needed to snapshot various disks (postgres, redis, etc) when sourcegraph.com was migrated to use https://github.com/sourcegraph/deploy-sourcegraph fork. We realized that we needed to manually specify a persistent volume for each deployment so that we could refer to the snapshot GCP disk. Unfortunately, this means that we had to manually create a GCP disk for each persistent volume, which is really tedious / error-prone.

https://github.com/sourcegraph/deploy-sourcegraph needs to either:

1. Be updated with instructions / stub PersistenVolume files that direct the user to create explicit disks for each deployment (which might not be true on non-GCP services)

2. Do more investigation to see how other people handle snapshots on k8s to see if there are more robust ways to accomplish this

3. Do more digging into how other projects do snapshots