Rewrite as lifecycle and spot monitors

[lox]

Thinking about #42 prompted me to try more elegantly splitting the spot and lifecycle monitoring into dedicated objects and leaving the Daemon entirely focused on process management.

This cleans up the SQS queue as soon as the lifecycle handler script has completed successfully, which seems about as quick as is possible.

Thoughts @itsdalmo?

[itsdalmo]

This is great stuff, really like the Monitor concept 👍

Some comments:

• I'd prefer if the handler was run inside daemon.Start() to ensure that we don't have to maintain the same code in two places, and also because handling the termination notice is not something I'd expect to be done by a struct with Monitor in it's name 🤔
• In the current code, the daemon will continue to run until both the LifecycleMonitor and SpotMonitor return, but really lifecycled should exit with an error if e.g. a queue is supplied but fails to create?

I think perhaps a combination of this and what I suggested in #42 could be neat?

[lox]

I made some of the changes you suggested, and drew some inspiration from #42. Lemme know what you think @itsdalmo.