Use NetworkX to identify bonds, angles, dihedral, impropers

[ahy3nz]

While messing around with using networkx to identify angles/dihedrals/impropers, some comparison functions are needed. This PR includes definitions of those comparison functions - I don't think these comparison funcs need to be super sophisticated, we just need to have some sort of way to compare and sort sites so the nx.graph can do something
I've attached an example jupyternotebook that does this. I think the idea of using some graph-matching techniques could be a robust way to identify angles/dihedrals/impropers, with some clear patterns that we can extend to other types of connections. As @volgy said, we may want to implement these graphs and matching algorithms in our own way #138
example.gz

[volgy]

Yes, this makes sense (need for any canonical ordering of sites to weed out duplicates found by the subgraph isomorphism). I still do not see crystal clear how the edge induced subgraphs will be found with the line_graph trick.

[ahy3nz]

I think most of the actual code is here - I tucked away all the functions in another module called connectivity.py, whose functions are just called by topology.enumerate_connectivity().

Angles: I think this is fairly robust - unit tests are on ethane, methane, a square, and a square with a diagonal. Do we need more unit tests for this?
Dihedrals: Code is written to identify dihedrals, waiting on #146, will need to modify existing unit tests or add some new omes
Impropers: Code is written to identify impropers, will need to modify existing unit tests or add some new ones

[uppittu11]

I done messed this up