celery should only run tasks for its deployment status

[btbonval]

By default, celery worker and beat both run all queues.

Restrict queues to staging or production because each runs different versions of code.

We don't want production messages to be processed by newer code of staging's tasks or vice versa.

forked from #407

[btbonval]

It looks like beat has no such queue switches. I guess beat talks to the local system and pushes tasks up to those mysteriously named queues for workers to execute? I don't really understand the differences in operation. I need a venn diagram and some powerpoint slides, instead there's a lot of source code.