Poor optimization around SETQ

[Wuerfel21]

As mentioned in #312, SETQ currently inhibits a bunch of optimizations by way of it being auto-marked as volatile. Though I'm not sure if that even covers all the potential troubles (esp since dead code can be disabled).

I know the block reordering functions used for the CORDIC optimizations (at least in theory) deal correctly with prefix ops. I think the only other places that might need to deal with it are the dead code eliminator itself, CanSwap and maybe the code that does hub read/write forwarding.

So probably not too much trouble?