Add configuration model generation

The configuration model takes a specified degree distribution and generates a random network with that exact degree distribution. It works as follows: give each node i k_i link stubs, where a k_i is the degree assigned to node i and a link stub is simply a link with only one end connected. Connect the link stubs to each other uniformly at random. This can result in networks with multiple links between nodes and self-loops. If any appear, just reshuffle the connections.

The nice thing about the configuration model is that it's so flexible. Generating a regular network becomes one line:

nw:generate-configuration n-values 100 [ 5 ]

Arbitrary scale-free networks can be generated like so:

nw:generate-configuration n-values 100 [ round (50 * (? + 1) ^ -1) ]

This is partially inspired by this SO question: http://stackoverflow.com/questions/25332514/assigning-neighbors-to-agents-in-symmetric-manner

Start by locating the extension entry point for nw:generate-configuration and the existing network-generation APIs. Implement the described degree-stub connection behavior, including reshuffling when self-loops or multiple links occur. Done means the example calls generate networks with the requested degree distribution without those connections.