Check if masses in OpenMM are canonical or edited a la HMR

[mattwthompson]

@mrshirts brings up a good point here, which I had not considered earlier. Interchange.from_openmm should make an effort to check that masses line up with elements. (I haven't had my coffee yet so I haven't thought of what sort of cases should be errors; I figure that something which appears to have had HMR applied should not error out, but also cannot have the non-canonical masses.)

Originally posted by @mrshirts in #742 (comment)

[mrshirts]

Maybe it should send a warning that the masses don't look right, rather than erroring and stopping?

[mrshirts]

There might not be a "true" mass, people might be doing an simulation looking at isotropic effects, seeing the difference of kinetics in O16 vs O18 water.