Add support for call sequences to the backwards implementation of frame
index elimination and convert X86 to use it.
The goal is to convert all targets to backwards frame index elimination
and backwards register scavenging, which is preferred because it does
not rely on accurate kill flags.
X86 does not actually scavenge registers during frame index elimination
but doing the conversion is useful for flushing out problems in the call
sequence handling.
I now think that this is just wrong, since operand(1) of ADJCALLSTACKDOWN is a completely different thing from operand(1) of ADJCALLSTACKUP.