Add some mitigation for server outages

[chreekat]

Stackage.org was down for hours, simply because I went to bed ~1 hour after finishing the migration to OpenCape, and didn't see that there was a problem until the next morning.

Yikes.

One mitigation is to give more admins access. This is already done.

It would be good to brainstorm some other mitigations. Watchdogs, for instance.

Related: commercialhaskell/stackage-server#345

[chreekat]

I think a simple systemd timer that tries to fetch a page from the service directly (via localhost:3000), and restarts the server on failure, would be a good start.