Allow configuration models to specify a minimum discrepancy between the sampled (di)graphical sequence(s) and the provided distribution

[ClaudMor]

Since the current algorithms, in order to sample a graphical sequence, may very well select a degree sequence that "has little to do" with the provided distribution, it should be useful to come up with a metric that accounts for such discrepancy and to give the user the possibility to specify a lower bound for such metric.