add molecule parsing functionality

[sfluegel05]

As noticed in ChEB-AI/python-chebai-graph#40, the SMILES parsing across different workflows is not consistent. While chebi_utils provides a sanitation function and a molblock parser, during inference, SMILES (or InCHI) strings can be used as input.

To standardize SMILES / InChI parsing across different repositories, I added a smiles_or_inchi_to_mol function.

The new SMILES / InChI / Molblock processing works as follows:

• Try to parse with the RDKit function (Chem.MolFrom*, sanitize=False, removeHs=False)
• If that yields None, return None with a warning
• Otherwise sanitize and return
• Sanitization tries to apply the update_mol_valences and Chem.SanitizeMol functions
• If that fails, it instead applies UpdatePropertyCache and FindFastRings to get the minimum properties needed for GNN processing