[mkldnn_matmul] enable mkldnn matmul for aarch64 bf16 devices (#83671) (#85546)
this PR enables mkldnn matmul for aarch64 bf16 devices for both bf16 as well as fp32 input.
This PR is dependent on
cpuinfo commit update PR: https://github.com/pytorch/pytorch/pull/83620
Issue: https://github.com/pytorch/pytorch/issues/83594
This is a reland of https://github.com/pytorch/pytorch/pull/83671
Pull Request resolved: https://github.com/pytorch/pytorch/pull/85546
Approved by: https://github.com/kit1980
