[llvm] [AArch64] Utilize `XAR` for certain vector rotates (PR #137629)

[David Green via llvm-commits]

davemgreen wrote:

This could to with some extra tests for fixed and scalable vectors, and to make sure the existing tests work. You might need to be careful about how the zero gets generated, I think it might need to generate a `MOVIv2d_ns 0`, but there are several ways to specify a zero vector and many of them are equivalent.

https://github.com/llvm/llvm-project/pull/137629