Non-intuitive behavior when the number of Wannier functions is different for the two spin channels

Currently the workflow does not complain for calculations which are spin-polarized and be number of Wannier functions (occ + emp) is not the same for both spin channels.

However, `kcp.x` expects both spin channels to have the same total number of Wannier functions.

Running a calculation with a mismatched number of Wannier functions should give an error and exit.