intel/fs: Cleanups and minor optimizations ahead of scalarization MR

I've been working on a branch to improve handling of convergent values in the Intel backend. There were some commits in that series that changed things in ways that should have not affected the final code generated, but there were changes. When investigating some of those surprising changes, I found that the same improvements could be gained other ways.

When I put that MR up, I really want the "this commit shouldn't change anything" commits to change as little as possible. The first 3 commits in are small cleanups and fixes, and the rest of the MR mostly implements those other small improvements.

Note: The results reported in this MR are on top of !27044 (merged). I don't think that will make much difference, but it is worth mentioning.