[MLton] Port to HP-UX

[Stephen Weeks]

> The mlton.share test also fails.  It's a bit more mystical to me.
> It reports bigger sizes than what the test expects.  Any suggestions
> on how should I start debugging this?

It's probably not a problem, and is also unrelated to the recent share
bug that only affected generational GC, which is unused on small
regressions.  The sizes of objects change from platform to platform,
so the sizes reported by MLton.size change too.  It would be
interesting to see the output though, just to confirm that nothing
strange is happening.  Could you send it?