[llvm] [AArch64] Improve lowering of truncating uzp1 (PR #82457)

[David Sherwood via llvm-commits]

https://github.com/david-arm commented:

This is looking good! I just had a couple of comments. For my comment about ignoring the input type and just always doing the transformation you can use your judgement about what's best. If you try my suggestion and it leads to worse code then feel free to ignore it! It just seems a bit odd that we care about the input type when someone has already created a UZP1 node, which essentially implies the type already.

https://github.com/llvm/llvm-project/pull/82457