Graph-ensembles

------------------
* Corrected distribution files for missing models

------------------
* Corrected sampling with ConditionalInvariantModel
* Added new weighted sampling method for aggregate graphs

------------------
* Added the sparse ConditionalInvariantModel
* Corrected issues with CremB sampling

------------------
* Improved and corrected support for the computation of average nearest neighbour properties and degrees.

------------------
* Added to the spark module a function to compute the confusion matrix elements at various thresholds of the probability matrix.

------------------
* Major update of graph classes into four new categories (Graph, DiGraph, MultiGraph, MultiDiGraph) in line with Networkx.
* Cleaned up models and organized in three modules (dense, sparse, spark) based on how the computations are performed and results are stored.
* Introduced better inheritance through Ensemble classes based on the newly defined graph classes.
* Added testing in spark and updated testing of MultiDiGraph Ensemble classes.