[llvm] [TableGen] Resolve GlobalISel strict weak ordering check. (PR #197071)

[Matt Arsenault via llvm-commits]

arsenm wrote:

> I'm trying to make sure I understand your comment correctly before merging. Do you mean: "the change in sorting due to this patch doesn't exactly match how the DAG orders patterns and updating the tests as this patch is doing makes sense"?

I mean ordering patterns with more predicates than less predicates I think makes sense in the abstract. I was just trying to map how the DAG orders patterns by complexity, and I'm not sure it's a 100% match but it doesn't really matter for the purpose of making this not assert 

https://github.com/llvm/llvm-project/pull/197071