Current implementation uses new_gfn2_calculator and a respective wavefunction to project guess charges on to that.
As this is in general only a poor workaround, it prevents the user also from using guess charges with Z>maxZ(xtb), e.g., using EEQ for obtaining actinoid charges.

Changes:

• General charge and atomic dipole moment array (spin=1) for driver_guess instead of the gfn2_calculator
• Renaming ceh_guess to ceh_singlepoint to be consistent with gfn1, ...
• Interface for EEQ and SAD guesses so that they can be used with partitioned shell charges or for raw atomic charges