[Portability] Better macOS support

[TLCFEM]

The binaries compiled on macOS via CI/CD works on the specific version of macOS with the specific version of compilers (GCC/Clang).

Only OpenBLAS is linked, which may not be optimal.
To unleash further performance, the support of MKL shall be enabled.
macOS is not likely to be run on Intel chips.

The following tasks need to be done:

• Improve portability by making the binary less dependent on external libraries.
• Explicit and clear instructions on how to execute the binary on different versions of macOS.
• Configure CI/CD to enable MKL.

[TLCFEM]

Start from v3.6, the arm64 build is distributed.
As I have no access to any arm64 machines, the compilation is currently emulated on amd64 machines.
The performance is bad.

Any help with regard to testing and benchmarking on arm64 is appreciated.