missing ECI because of cutoff issue

[zhongpc]

 max_lp = max(exp_struct.lattice.abc) / 2 
        for size, diameter in sorted(cutoffs.items()):
            new_orbits = []
            neighbors = exp_struct.get_sites_in_sphere([0.5, 0.5, 0.5],
                                                       diameter + max_lp,
                                                       include_index=True)

get neighbors in [0.5,0.5,0.5] with a max_lp can potentially miss the interactions with the centroid. For example, four-anion interactions on a tetrahedra.

To resolve this issue, by changing max_lp = max(exp_struct.lattice.abc) / 2 into max_lp = max(exp_struct.lattice.abc) would work.

This gives exactly the same correlation function as pyabinitio. (More general tests all also welcomed)

[lbluque]

This is the same as #104. Closing this one to have only one conversation.