Time bug work around

[Stephen Weeks]

> As to getrusage() being fixed some day, I had Rico test it on the latest
> kernel (2.4.6) and he could not get it to fail after many many hours.  Still,
> it will be a LONG time before everyone upgrades to kernels that fix the problem.
> I think the right thing is to have our own getrusage() which works around the
> problem.  It is trivial to write, and I am more than willing to do it.  It is
> true that it isn't `our' job, but note we did the same kind of hack when Linux
> had the bug that user/sys time distinction would broken on SMP machines.  It
> really is an easy work around.

OK.  Go to it.  I guess when we want to add fields to MLton.Rusage.t, we can
still use the hack for the time fields and calls to getrusage for the other
fields.