Support optimized kernels in top-level selective build

[GregoryComer]

Support optimized, portable_optimized, LLM, quantized, and backend-specific op libs in top-level kernels target selective build.

[GregoryComer]

#14203 adds support for optimized kernels. Quantized + LLM kernels may need more discussion. Landing the optimized kernels should be enough for GA.

[GregoryComer]

Closing as #14203 is merged.