Support automatic scaling of workers, based on queue length

[jonnor]

Should have hooks for, or provide ability to, scale number of workers up or down depending on number of messages in queue.
Algorithm should probably attempt to maintain a certain average processing time, bounded upwards by a price limit.

Heroku is the most interesting target right now, but support must be general
https://devcenter.heroku.com/articles/scaling
https://devcenter.heroku.com/articles/platform-api-reference#formation

[jonnor]

See https://github.com/jonnor/guv