Add internal error handling

[Olivia5k]

The distributed nature of this project makes for a hard time figuring out if something goes awry with the project itself. As such, we want them to report to specific points whenever internal errors (i.e. the script itself dying, not a test failure).

Sentry seems like the best fit as a general purpose solution.

Todo

• [ ] Add required sentry field to piperd.yml.
• [ ] Add Sentry code to agent launch.
• [ ] Add piper-specific data to the error report. Good start would be the piperd.yml and the currently running build configuration.

  Optional nice to haves

• [ ] Email logging if Sentry should be unavailable. Should be pretty simple with Logbook.

[Olivia5k]

Come to think of it, maybe Sentry actually has a fallback device like the nice-to-have built into it. Investigation would be useful!