Specifying a Limit For Parallel Proposed Runs Before Queueing

Right now Tracked runs can only run sequentially, while proposed runs will run in parallel on the same stack.

You can easily ignore multiple proposed runs inside of a push policy but this means that they are all dropped.

This is a concern when any number of proposed runs is less than the number of private workers available to a stack. With a central pool it can be totally saturated, it can also cause throttling on providers.

The solution would be allowing a user to specify a number of parallel proposed runs allowed before they queue.

Ideally a push policy update to match the existing proposed run logic.