The monitoring server needs Redis, but not the separate server that's
used for the breached-passwords bloom filter in dev/prod. This splits
that server out to its own state, so that it doesn't need to be set up
on the monitoring server.