constant switch test bug in x86Translate

[Stephen Weeks]

> O.k., it is easy enough.  I think I'll integrate this solution with the
> 3rd solution as well; I've never seen it happen in practice, but there's
> no reason why the Transfer.replace bug couldn't be raised by the
> copyPropagation phase under normal circumstances.

Sounds good.  When you get a stable codegen again, send it out and I'll try to
recreate the "mlton: toX86Operand: strange Offset" bug.