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

[Rajveer Singh Bharadwaj via llvm-commits]

Rajveer100 wrote:

I have made some changes. So, I feel like the tablegen pattern could be a little less strict, also I am not sure about converting `v1` to `v2` since we catch the pattern and its just best to emit the final instruction rather than converting it to `v2` again and then further lower in tablegen, let me know your thoughts.

The code has simplified significantly though.

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