Create `SimpleArray::matmul_veclib()` to wrap `?GEMM()` in macos Accelerate/veclib

[yungyuc]

To follow up the speeding up work issue #715 and PR #767, create a new member function matmul_veclib() that uses ?GEMM() in macos Accelerate/veclib.

Its performance should be benchmarked against the naive matmul() and the sped-up matmul_fast().

The new function matmul_veclib() only needs to work on Apple Silicon. Wrappers for other vendor libs will be follow-up work.

[ThreeMonth03]

@yungyuc I will solve this issue later.

[yungyuc]

@ThreeMonth03 When creating follow-up issues, please mention this one to relate them.