HOOMD Forcefield compatibility issues

[CalCraven]

HOOMD writer in GMSO is incompatible with forcefields that parameterize connections by a mix of types and classes.

Bug Report:

The majority of the forcefields we have in the ecosystem define the connection types by a set of classes that makes up their components. However, we technically do support defining the connection by types as well, to give users flexibility to define a dihedral specifically while generally defining other connections by their classes (see this PR discussion for more details). This can cause confusion when identifying unique sets of connection_types by a string that makes up the 4 members; there needs to be some flexibility to write by either atomtype or atomclass.

This issue will need to be tested by adding a forcefield that behaves this way. Something such as

DihedralType = class1, class2, class3, class4 
AngleType = type1, type2, type3

I think we also need some more specificity of how to define a forcefield i.e., is it acceptable to write an bondtype as:
BondType = class1, type2

I believe the plan will be to address this in the coming months

TLDR; to_hoomd_snapshot and to_hoomd_forces will not work if your topology is defined in such a way as to be have both type and class definitions for connections.