Evaluate the usage of Celery for a replacement for the evaluation service's `runner.py`

[christophertubbs]

The evaluation service uses a script named runner.py to determine when to launch an evaluation. This is hand written, works fine, and is set up to run on multiple machines. Despite this, Celery may be a better alternative since it's an official product developed to do exactly what runner.py does.

Investigate it and record if it is worth replacing runner.py with it. Add notes somewhere, either in this tickets or within the wiki.