[llvm] [AArch64] Fix load latency costs without a usable scheduling class (PR #223416)

[John Brawn via llvm-commits]

https://github.com/john-brawn-arm approved this pull request.

LGTM. It feels like probably we should be able to handle variant scheduling classes in some way (e.g. looking at the C1Nano scheduling model we should probably be able to default to the variant marked NoSchedPred), but this is an OK way to handle it.

https://github.com/llvm/llvm-project/pull/223416