Enable sleef, BLAS for executor_runner_optimized (#1354)

Summary:
Pull Request resolved: https://github.com/pytorch/executorch/pull/1354

This library will prefer optimized kernel, and use portable kernel as a fallback kernel.

For ios, we can also use sleef and blas library, with minor modifications.

Then we can deploy the runner library with executor_runner_optimized for ios. This improves ViT performance.

Reviewed By: kimishpatel

Differential Revision: D51748530

fbshipit-source-id: d31321219090599e3e061445cdbc5a953de88366