Cross-interacitons workaround

[mattwthompson]

Describe the behavior you would like added to Foyer
We have had much (#128 #140 and elsewhere in meetings) discussion (and a few incomplete solutions #180 #235 ) to the problem of assigning cross-interactions in foyer workflows. We intend to have an elegant solution with the new backend, but for the meantime, I propose a simple workaround.

Describe the solution you'd like
A function that takes in a parmed structure, some information about a custom cross-interaction that is to be added (sigma/epsilon, atomtype1, atomtype2, etc.) and adds it to the nbfixes in the structure. This would be a sort of foyer.utils function and it would require an extra step after the atomtyping process. The benefits of this approach are twofold: first, it works without needing the backend implemented (a.k.a. it can work now) and second, it is a simpler solution than mirroring all of the analogous OpenMM internals.

Describe alternatives you've considered
In a separate project I have this sort of function implemented, but not for general use. It works well enough for that purpose, which includes custom cross-interactions. @zijiewu3 and the UD groups may have implemented something similar for their DNA systems (if not, this will solve their issue)

Additional context
As it's a band-aid, this should be deprecated after the backend is fully integrated.