Stacksize runtime bugs.....

[Daniel Wang]

I think, there a runtime bug associated with the stacklimitcheck and/ or
stack resizeing. As my code seem to segfault even though it gets all the way
through and typechecks all the way... I had seen ths problem before but, I
think Stephen set me a fix a while back, but that fix may not have been
fully general. If I diddle with the stacksize resize policy this problems goes
away.  It also could be miscommputing "maxframsize" parameter.

I'll see if I can find a simple example, that causes this problem. 

Also, I tried running mlton with a fixed stacksize by make the
initialStackBytes function return something appropiate, but it didn't seem
to work. Can I get a hint as to how to start the runtime with a large fixed
stack?