Investigate using deployment instead of stateful set in helm chart

[tpoliaw]

With a stateful set, the service has to be taken offline while updating to a new version. SQLite is meant to be able to deal with multiple processes accessing the same DB location so it should be able to cope with two instances of the container running during rolling updates.

Things to check

• Does sqlite's file locking work in whatever filesystem version k8s uses? The docs flag NFS mounts as being problematic
• Can multiple containers bind to the same PV? We're reusing the same PV for all instances of the container to preserve state from one process to the next. If there are two instances of the pod at any point, they'd need to access the same file.
• Does using a deployment/single instance replica set actually make the service uninterrupted during updates? If it doesn't there's no point updating from what we currently have.