masOS rsqrt failure

[kennykos]

The simple program

import pykokkos as pk

@pk.workunit
def workunit(i):
    j: float = rsqrt(i)

pk.parallel_for(10, workunit)

fails with

In file included from <path-to>/pk_cpp/tmp/tmp_workunit/types_fa7153f7ed1cb6c0fcf2ffb2fac21748/OpenMP/bindings.cpp:10:
<path-to>/pk_cpp/tmp/tmp_workunit/types_fa7153f7ed1cb6c0fcf2ffb2fac21748/OpenMP/../functor.hpp:5:301: error: use of undeclared identifier 'rsqrt'; did you mean 'sqrt'?
    5 | template <class ExecSpace>struct pk_functor_workunit{struct workunit_tag{}; pk_functor_workunit(int32_t pk_field, int32_t pk_randpool_num_states, int32_t pk_randpool_seed){  }; pk_functor_workunit(int32_t pk_field){  };KOKKOS_FUNCTION void operator()(const workunit_tag& , int32_t i)const{ double j= rsqrt(i); };};
      |                                                                                                                                                                                                                                                                                                             ^~~~~
      |                                                                                                                                                                                                                                                                                                             sqrt
/Library/Developer/CommandLineTools/SDKs/MacOSX.sdk/usr/include/math.h:457:15: note: 'sqrt' declared here
  457 | extern double sqrt(double);
      |               ^
1 error generated.
make[2]: *** [CMakeFiles/kernel.cpython-313-darwin.dir/bindings.cpp.o] Error 1
make[1]: *** [CMakeFiles/kernel.cpython-313-darwin.dir/all] Error 2
make: *** [all] Error 2

CMake build in pk_cpp/tmp/tmp_workunit/types_fa7153f7ed1cb6c0fcf2ffb2fac21748/OpenMP failed

on macOS.

Fixes:

• fix rsqrt on macOS
• update macOS CI to run all tests on macOS (currently, only on example is checked)