Add --drain flag to worker command

[j4mie]

python manage.py worker --drain will start a worker process which will drain all jobs in the queue (if there are any) and then exit. This addresses #16 (from seven years ago 😱) and #52, and is an alternative (and much simpler) solution to that proposed in #56.

[jordaneb]

This is a nice simple approach to solving this problem but I think we should warn that if the queue exceeds a certain size then the cron job could run for a really long time and depending on the frequency of the cron job then multiple DBQ workers could be spawned at once.

Something along the lines of: "Note that if the queue is large it will be necessary to manage the lifetime of the process with a tool like timeout (http://man.he.net/?topic=timeout&section=all) to make sure it doesn't run forever or that multiple worker processes aren't spawned depending on the frequency of your cron schedule."

[collinr3]

Thanks @j4mie. I agree that it is simple and that there is attraction in that.
I'd previously gone for the more complex #56 --shift_limit approach, specifically because of the potential issue @jordaneb raises. The dynamic between --rate expected volumes of jobs, average job execution time and CRON interval is a complex one. The unknown is what undesirable side effects may be introduced into the data should there be multiple workers processing the same queue. That said, the issue cannot be easily mitigated in a CRON, environment, whatever the approach, so the 'warning' may be the most pragmatic.
One comment on the test for --drain consider adding an extra step to submit a further Job after the worker has initially drained the queue to check that it doesn't get processed.