Implements loop distribution transformation

[kaushikcfd]

TODO:

• Do index analysis to verify the validity of the iname-duplication passes.
• Add complicated regressions.

Draft because:

• includes commits from generate_loop_schedule_v2 #350.
• Needs pytools==2022.2 (Compute topological sort with dynamic key pytools#145)

[kaushikcfd]

This is on hold currently as @inducer and I agreed in a personal chat that:

1. The soundness checking is hokey and must wait for proper dependencies in loopy.
2. The greedy approach of selecting the duplication groups doesn't achieve minima and even in practical cases does not get a reasonable solution.

[inducer]

The soundness checking is hokey and must wait for proper dependencies in loopy.

To the extent that this is @kaushikcfd channeling what I said, I feel I should reword this. 🙂 This PR introduces special-purpose soundness checking for the case of loop distribution. While that's clearly useful and important, I think that this could be (more precisely, and at the same in a non-special-purpose manner) handled by generic dependency checking along the lines of #479. That's something I'd like to make some progress toward this summer.