Build optimized library with CMake (#2530)

Summary:
Support optimized kernel library in CMake builds.

Overriding pre-existing CI failures in llama runner and mac unit test.
bypass-github-export-checks
bypass-github-executorch-ci-checks

Pull Request resolved: https://github.com/pytorch/executorch/pull/2530

Test Plan: Built and ran example models using combined optimized+portable target. CI is clean, barring pre-existing failures. Also, confirmed that optimized kernel is being used by running add example and adding temporary logging to optimized kernel.

Reviewed By: kimishpatel

Differential Revision: D55118200

Pulled By: GregoryComer

fbshipit-source-id: b094980e9fc402e316a4d7cfb24ee3646a00d64e