Handle backend timeout gracefully

[giftig]

Not currently logging anything when it takes a long time to talk to redis; the request times out and spray tells the client it wasn't able to respond in a timely manner, but nothing is logged to help narrow down the problem. It'll be useful to log before and after hitting the db so that timings are visible, and to handle timeouts by logging the issue and possibly serving a JSON response to match the rest of the API rather than falling back on spray's text/plain one (maybe. Status code is probably good enough; passing back JSON won't add much information).

[giftig]

I should also document the dangers of sharing a redis db with flippy since it SCANs the keyset.