[llvm] [AArch64] Use `AArch64ISD::UADDLP` for manual widening adjacent arithmetic (zext/shuffle combination) (PR #189255)

[David Green via llvm-commits]

https://github.com/davemgreen commented:

High level comment - how much of this could be handled with a tablegen pattern and maybe a canonicalization of one form to the other?

Or otherwise, is it simpler to match from AArch64ISD::UZIP1 and AArch64ISD::UZIP2 instead of the shuffles?

https://github.com/llvm/llvm-project/pull/189255