Add queue for computing custom cohorts

[mitalia]

When a user edits a cohort, the change should be queued for calculation. Cohort calculations ideally should run on a server other than the main web server, but at a minimum should not run immediately to prevent mistakes or rapid edits from clogging things up.

Once a user edits a cohort, the allele frequency computation should be put into a queue. It would be good to do a quick diff to see what has happened to the cohort. If users have only deleted samples, these changes can be immediate, with the new cohort simply being a copy of the old, with the samples in question removed.

Otherwise, the entire cohort should be recomputed. Ideally, it would be good to delay this just slightly, so that people have a chance to edit before the computation starts. We'll have to see how long these take to determine what frequency they should run. Ideally, users wouldn't have to wait too long for them. But we also don't want to have the database churning computing all sorts of cohort frequencies at once. I am also cognizant of not over-complicating this, so if there's a simple solution, I'm game.