Painfully slow Stdlib build needs fixing

Once upon a time, there was a unit test bug that caused me grief:

. Some time ago, those apex unit tests (#2278) ran clean. They no longer do so. I was going to use our dear friend, git bisect, to see if I can determine when that unit test code ceased to work.

However, the fact that it can take 45 minutes to compile the Stdlib makes use of
this tool utterly impractical, unless you can come up with a failing example that
does not require Stdlib. The work required to develop said failing example can take
days or weeks. This is NOT a good use of our time and energy.

The critical path in a Stdlib build is, based on observation while waiting for things
to complete, long, much like a strand of spaghetti. We could get, I think, a
factor of two speedup by building all targets concurrently (e.g., seq and mt_pth).
I don't have any other bright ideas. Or other kinds.

I think the seq/mt_pth change "should" be simple, but there is presumably something that makes it non-simple.