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

[David Green via llvm-commits]

https://github.com/davemgreen commented:

Thanks - this looks good. Can you add tests like `xar_instead_of_or` but with types `<4 x i32>`, `<8 x i16>` and `<16 x i8>` to llvm/test/CodeGen/AArch64/xar.ll, and tests like `xar_nxv2i64_l_neg2` but with types `<vscale x 4 x i32>`, `<vscale x 8 x i16>` and `<vscale x 16 x i8>` to llvm/test/CodeGen/AArch64/sve2-xar.ll. Not all of them are expected to transform, but we should make sure we test all the combos.

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