[llvm] [AArch64][NEON] Fold insert(zero, extract(X, 0), 0) -> X when X is known to zero lanes 1-N (PR #213940)

[via llvm-commits]

================

----------------
CarolineConcatto wrote:

The reason I did not add yet the FP is because I am not 100% sure I could assume the same behaviour for the floating point.
This is the operation for ADDV=>V{esize}(d) = IntReduce{esize, datasize}(ReduceOp_ADD, operand);
While this is the operation for FADDP=> V{esize}(d) = FPAdd{esize}(element1, element2, FPCR());

https://github.com/llvm/llvm-project/pull/213940