Come up with an automated end-to-end BonnyCI test

[gandelman-a]

Validating the BonnyCI system is functional from an end-to-end, user POV, I often find myself:

• Creating a fork of a BonnyCI-managed sandbox
• Committing a change to that fork
• Proposing that change via a PR from the fork to the BonnyCI-managed sandbox
• Ensuring check jobs run, pass/fail as expected, and publish logs to the correct place
• Getting someone to approve the change
• Ensuring gate jobs run, pass/fail as expected, and publish logs to the correct place
• Ensuring the patch lands in the upstream BonnyCI-managed sandbox repo

This should be distilled into an automated test case that can test various scenarios of the above. There are some problems to be figured out about how we manage github credentials, sandbox setup, etc. but for the first iteration it might be OK to just lean on a test config that contains info for pre-existing github users and repositories.

The long term goal with this is to a have a small test suite that is portable between BonnyCI environments. In theory we should be able to point it at any repos and let it run, and validate that the BonnyCI environment managing those repos is functional. This can be used in CI and QA to validate test environments, in operations to validate a new production environment does what its supposed to, and used in monitoring to periodically test the health of a BonnyCI environment. Note that this is not a test of hoist itself, but of the running thing that hoist produces.