RSSA backend

[Matthew Fluet]

> Matthew, have a go at getting it to work with the x86-codegen when you
> get the chance.  The main problems you may encounter that I can think
> of are the new stack pointer convention and the liveness info in the
> Machine IL.  Let me know if you have problems or when it's ready and
> I'll merge RSSA in to the main line.

What liveness information is being stored with MACHINE blocks?
I see stuff like:
L_25 {kind = Handler(0), live = (SU(4), RP(14))}:
    RP(14) = globalpointerNonRoot [0]
    RP(15) = RP(14)
    Goto L_124

where I wouldn't expect to mark RP(14) as live into L_25?