Time bug work around

[Matthew Fluet]

> We still have the problem of the unexplained Time exception in my ML test code.
> Matthew suggested that it could be in the Time calls in rusage.sml, but I
> don't see how the out-of-order aspect could do that.  On 2.2 kernels I saw
> the microsecond field = 10^6 some times, but I don't see how that could do it
> either.

Looking through the code, the only things that could raise the Time
exception were Time.- and some of the Time.from* functions.  I still think
it has to be coming from a Time.from* call within the MLton.Rusage.rusage
call, because if it were coming from showTime, you would have seen "Binfo
for ..." before the unhandled exception was raised.