Improve handling of errors.

[keith-turner]

Currently when an Observer fails an exception is logged and Fluo retries when it sees the notification again. What if anything can be done to improve this?

Below are some possibilities.

1. Retry failed notification less frequently.
2. Have an API and command line tool for finding failed notifications.
3. Document via a blog post how to use log stash with Fluo to find failed notification log messages.