Support resonance forms of residue definitions

[Yoshanuikabundi]

For example, the CCD dictates that the positive charge on Arginine (and the double bond) should be on NH1, but a PDB file could place it on NH2 (see, eg, 5EIL A:ARG70). At the moment, this causes the residue to fail to load, as there is a charge mismatch. We could define a new residue definition that covers this variant, but this would introduce an ambiguity when the PDB file defines no formal charges at all (for example, just about any arginine produced by OpenMM/PDBFixer).

This is probably blocking #18, as otherwise normalization could disrupt residue matching.

The solution may be to replace topology_from_pdb's call to ResidueMatch.agrees_with, which is used to allow multiple matches to proceed if the assigned chemistry is unambiguous, with a normalization procedure. If all matches assign the same chemistry once normalized, that normalized chemistry is assigned. Common resonance forms would then have to be manually patched in to the CCD cache.

[j-wags]

For the record: I believe @Yoshanuikabundi and I spoke about this and we decided in these cases that we should trust the CCD and disregard the charges in the PDB file.