intel/fs: Housekeeping in cmod propagation with saturate

Back in a7957009 we thought that the conditions were generated before the clamping for the saturate was applied. We quoted some documentation to support this, but the documentation was misleading. In !3348 (merged) @yk performed some experiments that prove this is not the case. The conditions are generated after the clamping for the saturate.

This MR removes all of the extra restrictions on cmod propagation into an instruction with a saturate modifier. There are some shader-db improvements, but this MR is more about deleting code.