Multiple graphs dataset example

[Gregg Kellogg]

Spec doesn't have an example for running against a dataset.

From #111:

As far as I can tell, all examples only have the default graph in the dataset. With quads, the graph is just one more endpoint of the edge, so probably no increased complexity as far as cases in the algorithm are concerned, but it looks a bit strange that it speaks about quads when throughout there are just triples in the examples (especially when serialised for hashing).

I find myself wondering how hard the algorithm would have to determine that a Möbius ladder graph is not isomorphic to a circular ladder graph of same order. (For graph isomorphism to appear as a subproblem of c14n, consider a dataset with two graphs whose names are blank nodes. At what point can it tell those nodes get different hashes?) Or for that matter, Möbius ladder to a circular ladder of order 2 more. The algorithm does not directly use multiplicity of hashes, does it? Only whether hashes are unique or not. Hence a 10-vertex 3-regular graph is not so obviously distinct from a 12-vertex 3-regular graph.