[llvm] [AArch64][SME] Preserve Chain when selecting multi-vector LUT4Is (PR #161494)

[Benjamin Maxwell via llvm-commits]

MacDue wrote:

Looks like the only other case was the non-lane variant of `lut4i` (which I've fixed now too). I've moved the tests to a precommit (https://github.com/llvm/llvm-project/pull/161505) to make the changes more apparent. 

https://github.com/llvm/llvm-project/pull/161494