This commit implements a new capasity to human readable string function
that supports all kinds of speeds ranging from 1k to 1000Exabit++

  * Use direct dictonary lookups instead of get_edge where possible.
  * Rewrite all metrics to be edge dictionaries with key=weight.
  * Use nx.average_shortest_path_length instead of homebrew version.

Conflicts:

	utils.py

Morten Knutsen authored Sep 14, 2009 and Morten Knutsen committed Sep 14, 2009

Use .edges(data=True) where applicable.

but is somewhat slow still.