raise error when atom name encoding is violating the expected dimension

[wckdouglas]

In this PR, we are aiming at raising an Error when an invalid atom name is generated from the given smiles.

When a larger molecule (this example) is feed into boltz, we encountered a SI101 atom name where it violates what we expected (4-char string) in convert_atom_name. test code:

from rdkit import Chem
from rdkit.Chem import AllChem

seq = "CC(C)[C@@H](C(=O)C1=NC2=C(O1)C=CC(=C2)CO[Si](C)(C)C(C)(C)C)NC(=O)[C@@H]3CCCN3C(=O)[C@H](C(C)C)NC(=O)OCC4=CC=CC=C4"
mol = AllChem.MolFromSmiles(seq)
mol = AllChem.AddHs(mol)

# Set atom names
canonical_order = AllChem.CanonicalRankAtoms(mol)
for atom, can_idx in zip(mol.GetAtoms(), canonical_order):
    name = atom.GetSymbol().upper() + str(can_idx + 1)
    if len(name) > 4:
        print(name)
    atom.SetProp("name", name)

The above molecule gives a SI101, and convert_atom_name returns a (51, 41, 17, 16, 17) which later numpy complains not being in the correct dimension.