Add arm64 (armv8) kernels to all pair-counters (targeting Apple M1 and above)

[manodeep]

Is your feature request related to a problem? Please describe.
Add armv8 kernels to all pair-counters

Describe the solution you'd like
While Corrfunc now compiles and runs on arm64, at runtime only the fallback kernels are available. Ideally, we should have armv8 kernels to all pair-counters

Describe alternatives you've considered
N/A

Additional context
Based on my initial implementation, the armv8 kernels speed up the computations significantly on my M2, compared to the fallback kernels. However, that work stalled because there was no CI to test any implementation regularly and I didn't want to keep working on it without a working CI. Now, it looks like that the arm64 runners are available with macos-latest - so the CI should be able to handle testing the code.

But this also means that we need to add an additional CI OS to check for Intel macos