RSSA backend

[Matthew Fluet]

> > Why am I plagued with inexplicable regression errors?  I can't reproduce
> > these either locally or on the machine where the rest of the regressions
> > are still going (i.e., over on that machine, I'm using exactly the same
> > executable and regression files that yielded the above).
> > 
> > Alright, it only seems to crop up when running in the regression script.
> 
> I just checked out the main trunk and used an NJ compiled G0 to run
> the regression script on {array,callcc,fft,filesys,world1} without
> problems, on all three of our linux machines here.  I have no idea
> what's going on.

Yeah.  I'm not sure either.  I suppose it could be a bug in the compiler I
used to compile the main trunk; that would have been pretty much
equivalent to the pre-rssa-merge tagged code.