aco: Improve SCC no-compare optimization
The two ACO commits pulled out of MR !13155 (merged) - the main purpose of these two optimizations is to 'mitigate' a minor regression from that MR.
Fossil DB stats of the whole MR on Navi 21:
Totals from 70665 (52.38% of 134906) affected shaders:
CodeSize: 248742480 -> 247858588 (-0.36%)
Instrs: 46959475 -> 46738502 (-0.47%)
Latency: 392326885 -> 392224589 (-0.03%); split: -0.03%, +0.00%
InvThroughput: 67707786 -> 67703937 (-0.01%); split: -0.01%, +0.00%