add run-policy to cosmic-swingset

[warner]

What is the Problem Being Solved?

Once #3459 is done, we hope to have an idea of what sort of computron limit to apply to each block. Once #3580 has landed, we'll have a way to apply it. At that point, the task will be to update cosmic-swingset:

• create a computronCounter run-policy object at the beginning of each cycle
• use c.run(policy) instead of the current for() loop that calls c.step() some maximum number of times

For the solo machine, we can either leave the existing mechanism in place (drain the queue each time), or add a wallClockWaiter to limit the work-until-commit duration to something like 2-5 seconds.

Test Plan

I think this will have to be tested manually, with a load generator cranked high enough to cause the run-queue to remain full for a while.

[dckc]

Once #3459 is done, we hope to have an idea of what sort of computron limit to apply to each block.

Examination of one busy block suggests an 8-10M computron limit.

[dckc]

@warner you mentioned making the threshold configurable via governance. Is that required?

p.s. no, not required, from discussion with @michaelfig.