[llvm] Reapply "[SPARC] Use umulxhi to do extending 64x64->128 multiply when we have VIS3" (#135897) (PR #136475)

[via llvm-commits]

koachan wrote:

> Yes, but the pattern is the same, no? The original commit did pass pre-commit checks, but failed on a buildbot due to changed instruction order. Why was the order different between CI / build bot?

So I tried a bisect and seems like between the time I created the old PR and it getting landed, commit [74e8f29](https://github.com/llvm/llvm-project/commit/74e8f29f3138431abc9b393579733f3bbd0d85d8) landed and that changed the instruction ordering.
Not sure why the CI status on my PR doesn't get updated, though.

As for this PR, I think a rebase before landing it would be enough to shake out those issues?

https://github.com/llvm/llvm-project/pull/136475