[llvm] [AArch64][SVE] Prefer SVE over scalarization for masked load/store of bf16 vectors (PR #210986)

[Paul Walker via llvm-commits]

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

Looks good to me. While there has been significant improvement for scalable vector bfloat operations, this PR still comes with risk.  However, any issues triggered by this PR are likely things we need to fix anyway so I'm happy to accept that risk.  That said, given the release process has started I will not support a backport request.

https://github.com/llvm/llvm-project/pull/210986