Add support for loading partial charges from SDF tags using new standard

[j-wags]

The OE and RDK maintainers have agreed on a spec for storing partial charges in SDF. We should implement support for I/O of this data.

[j-wags]

Test suggestions from John:

(1) including a set of charged SDF files in our test dataset
(2) ensuring openforcefield.topology.Molecule can write charged SDF files via either RDKit or OpenEye
(3) adding tests of round-trips in all relevant directions between charged SDF files, OpenEye, RDKit, and Molecule?

I think this is exactly how we should go about this.

[j-wags]

Some notes as I move towards implementing this:

We will assume that partial charges should be written to a > <atom.dprop.PartialCharge> block in SDF files. Units for these values are assumed to be simtk.unit.elementary_charge (I'll consider whether we should explicitly specify this in another SD field)

It will generally follow the spec laid out here: rdkit/rdkit#2297

[j-wags]

Note: Keep a record of what we end up calling the partial charge field in the table in #135

[jchodera]

We should be sure to document (and version) the standard SDF tagging scheme for charges clearly in a readily accessible location so that it will be clear to other cheminformatics packages what the standard actually requires.