Openbabel 3.1 has a mistake in generating radical molecules from XYZs

[xiaoruiDong]

Openbabel 3.1 has a mistake in generating radical molecules from XYZs. This influences TS-GCN notebook that uses openbabel as the backend to perceive XYZs for radicals. A temporary fix here may be

1. checking the total spin multiplicity and valence of each atom
2. manually assign the multiplicity according to 1.

[xiaoruiDong]

openbabel/openbabel#2392

[xiaoruiDong]

This is a bug possibly due to a bad compilation of the anaconda binary (maybe only for macOS). Currently workaround for N, C, H, and O is added.