Add queue support about zimfarm

[automactic]

@kelson42 What do you think are some appropriate levels of the queue? How about:

Queue	Expected File Size
tiny	<=100MB
small	100MB-1GB
medium	1GB-10GB
large	>10GB

We can use this queue configuration as a start and adjust later.

[kelson42]

@automactic looks good to me

[kelson42]

Even if I wonder if it would not be simply better to give info about HW resources need at the time of the job scheduling and then start the docker worker with parameter specifying the HW limitation of the worker. Then the rest would run automatically with one queue (and filtering).

[automactic]

Celery does not work like that. If we have only one queue, then all free workers in the queue will get task evenly. A worker cannot say, let me inspect this task and if I don't like it then I will send it back.

Also, please define HW resources and limitations and how it is different from queue based solution.

[automactic]

This feature is being implemented in:

• #82
• #85