Improve PDB formatting with incomplete Monomer info

[fwaibl]

Introduction

By default, the PDB writer uses the residue name UNL, and sets very reasonable atom names:

>>> from rdkit import Chem
>>> mol = Chem.MolFromSmiles(r"ClCl")
>>> print(Chem.MolToPDBBlock(mol))
HETATM    1 CL1  UNL     1       0.000   0.000   0.000  1.00  0.00          CL  
HETATM    2 CL2  UNL     1       0.000   0.000   0.000  1.00  0.00          CL  
CONECT    1    2
END

We can also set the residue info manually. If this information is complete, everything works:

>>> at = mol.GetAtomWithIdx(0)
>>> res_inf = Chem.AtomPDBResidueInfo()
>>> res_inf.SetName("CLA ")
>>> res_inf.SetResidueName("CL2")
>>> res_inf.SetIsHeteroAtom(True)
>>> res_inf.SetResidueNumber(1)
>>> at.SetMonomerInfo(res_inf)
>>> print(Chem.MolToPDBBlock(mol))
HETATM    1 CLA  CL2     1       0.000   0.000   0.000  1.00  0.00          CL  
HETATM    2 CL1  UNL     1       0.000   0.000   0.000  1.00  0.00          CL  
CONECT    1    2
END

However, if the residue info is incomplete (e.g., we set the residue name but not the atom name), the PDB output will be invalid. Most importantly, the line is too short, so that the later columns get into the wrong place:

>>> mol = Chem.MolFromSmiles(r"ClCl")
>>> at = mol.GetAtomWithIdx(0)
>>> res_inf = Chem.AtomPDBResidueInfo()
>>> res_inf.SetResidueName("CL2")
>>> at.SetMonomerInfo(res_inf)
>>> print(Chem.MolToPDBBlock(mol))
ATOM      1  CL2     0       0.000   0.000   0.000  1.00  0.00          CL  
HETATM    2 CL1  UNL     1       0.000   0.000   0.000  1.00  0.00          CL  
CONECT    1    2
END

This happens if either the atom name or the residue name is missing. But IMO, the more important case is the missing atom name, since enumerating the atom names is more difficult that setting a dummy residue name.

Changes

• This PR changes the code so that default atom names are generated if they are not present in the AtomPDBResidueInfo object. The same code is used as in the case without any residue info, and I refactored that code into a new function GetDefaultAtomNumber.
• I also added an std::setw(4) before the atom name and an std::setw(3) before the residue name, so that additional spaces will be printed before the atom or residue name if it is too short. (See below: I should still add a unit test for that). Again, this avoids incorrect column alignment. If the user specifies a short name like "C1", it will still not be aligned properly within the column, but the result will be parsed correctly by PyMol. (PDB specifies that the alignment of atom names depends on the length of the element symbol. To do that perfectly, we would have to parse the user-specified atom name).

To do / Questions

• I did not yet make a unit test where the atom name is set but the residue is not. I specified the minimal length in the code, but did not yet test if it works. I will add another commit for that.
• I added a function called GetDefaultAtomNumber, since that code is now used in two different places. I only declared and defined this function in the PDBWriter.cpp file, since I don't believe that it can be useful anywhere else, and since there is no corresponding .h file. Is that ok?
• What do you think of the alignment issue? I believe that adding spaces is an improvement, but it is still not perfect. However, the user can simply supply their atom names with the correct alignment.

What do you think of this? Thanks in advance for your help :-)