[llvm] [AArch64][TTI] Fix extract cost for scalar fmul users (PR #212739)

[Anutosh Bhat via llvm-commits]

anutosh491 wrote:

The `reordering-single-phi.ll `test failed after the first commit because AArch64 now correctly produces different IR from x86: it reuses a lane from the vector load instead of emitting a separate scalar load. Updated the test to use target-specific check prefixes.

https://github.com/llvm/llvm-project/pull/212739