Feature request: -async

[potash]

There is a disparity between the doc and the implementation on asynchronous job execution. The doc describes an async option whereas the implementation is the command line option --jobs.

In my pipeline I have some steps that are bigger than others. For the smaller steps I want to run drake --jobs=16 but the big steps will run out of memory or fight for CPU (because the steps are themselves multi-threaded).

The docs describe the "async" option and I think it would be very simple and effective to simply allow "-async" to turn off async execution for a step. In that situation I would expect drake to run that step alone and not in parallel with other steps. In order to do so it may have to wait until previous steps are done.

(There are more complicated variations on this where we might have step options like jobs:4, indicating that this step needs/wants 4 job slots, etc. Another possibility would be for drake could make a $JOBS environment variable available to the step representing how many of the command line arg --jobs slots are currently available. But I think a simple -async would be enough for many use cases)

Thoughts? Thanks.