Use target slot analysis to simplify insert-rotate

[j2kun]

Rebased off #526

This gives us a cleaner implementation of the target slot selection, and has a few benefits:

• Now programs like hamming_distance and the simple_sum from Achieve the optimum rotations for insert-rotate on a simple_sum example #520 materialize their reductions as a sequence of shift-by-1 rotates, which means we can augment rotate-and-reduce to support replacing that with a logarithmic number of rotations. In particular, this increased the number of rotations of the hamming_distance test case by one, which is temporary.
• I found a way to incorporate the result of a dataflow analysis into DRR rewrite patterns, which is demonstrated in this PR by first running the dataflow solver, then walking the IR and attaching the output of the analysis as attributes to the relevant ops, and adding a helper NativeCodeCall in DRR to extract the attribute. I found out you can also use this to apply constraints (i.e., apply this pattern only if the op has the relevant attr). This method seems like it would be limited to simple data output by the analysis, not something complex like a struct or piece of a larger data structure.

[AlexanderViand-Intel]

• I found a way to incorporate the result of a dataflow analysis into DRR rewrite patterns, which is demonstrated in this PR by first running the dataflow solver, then walking the IR and attaching the output of the analysis as attributes to the relevant ops, and adding a helper NativeCodeCall in DRR to extract the attribute. I found out you can also use this to apply constraints (i.e., apply this pattern only if the op has the relevant attr). This method seems like it would be limited to simple data output by the analysis, not something complex like a struct or piece of a larger data structure.

Ohhh, nice! That's a cool little insight, independent of the main point of the PR/pass. We might want to start collecting these things and other "best practices"/"tips and tricks" somewhere so new contributors can discover them?

[j2kun]

Rebased and ready for any final review comments.

[AlexanderViand-Intel]

Looks good to me! Sorry for the delay, I thought I had reviewed this on Friday before leaving for Toronto, but apparently forgot 🙈