Error recovery strategies

[brenns10]

It is currently possible for the bot to (almost) silently crash. The only evidence of this is the fact that the slacksoc user is no longer "online" according to Slack. We can do a number of things to avoid this.

• Recover from panics at top-level function. I don't know if this will do much, because I'm sure that Go already outputs as much information as you could want when a panic crashes a program.
• Recover from panics at plugin level, at least long enough to log them.
  • it could even be possible to make plugin registration reversible, and then unregister any plugin which causes a panic, to avoid further damage
• Catch errors that are returned by plugin functions, and log them.
• Add a logging hook that sends error/critical messages to slack (if possible), and maybe an email