Refactor macOS CI

[zcbenz]

Refactor the macOS CI to build binaries and run CPU tests on the free GitHub-hosted runners, and then transfer the binaries to our self-hosted mac runner and run GPU tests there.

This work-arounds the extremely long wait time of our self-hosted mac runner, so it is only used once per each job and each run only takes 10 minutes.

This also work-aorunds the crash of MPI tests on self hosted mac due to the hardware being too new: openpmix/prrte#2416

  [ip-240-65-229-151:06579] *** Process received signal ***
  [ip-240-65-229-151:06579] Signal: Segmentation fault: 11 (11)
  [ip-240-65-229-151:06579] Signal code: Invalid permissions (2)
  [ip-240-65-229-151:06579] Failing at address: 0x0
  [ip-240-65-229-151:06579] [ 0] 0   libsystem_platform.dylib            0x0000000190097744 _sigtramp + 56
  [ip-240-65-229-151:06579] [ 1] 0   libprrte.3.dylib                    0x0000000102c7fef4 construct_range + 36
  [ip-240-65-229-151:06579] [ 2] 0   libprrte.3.dylib                    0x0000000102c7fc24 prte_hwloc_base_get_topo_signature + 632
  [ip-240-65-229-151:06579] [ 3] 0   libprrte.3.dylib                    0x0000000102ca21fc rte_init + 1940
  [ip-240-65-229-151:06579] [ 4] 0   libprrte.3.dylib                    0x0000000102c4bcd0 prte_init + 1044
  [ip-240-65-229-151:06579] [ 5] 0   libprrte.3.dylib                    0x0000000102c61228 prte + 3804
  [ip-240-65-229-151:06579] [ 6] 0   dyld                                0x000000018fccfe00 start + 6992
  [ip-240-65-229-151:06579] *** End of error message ***
  /Users/runner/actions-runner/_work/_temp/aa5791fb-4345-4dd7-b597-e55470c3f5d0.sh: line 2:  6579 Segmentation fault: 11  mpirun --bind-to none -host localhost:8 -np 8 -x python python/tests/mpi_test_distributed.py