[llvm] [AArch64][NeoverseV1/V2] Fix ASIMD FP convert and round scheduling (PR #224046)

[Asher Dobrescu via llvm-commits]

https://github.com/Asher8118 approved this pull request.

I've looked over this and ran some unrolled loops of FP converts on a V1 machine. The change looks good to my eyes. I've got one small nit pointed out in the diff, but otherwise LGTM.

https://github.com/llvm/llvm-project/pull/224046