Implement weighted and unweighted metrics (AUC, Matthews correlation coefficient, ...)

[rogerberm]

Averaging individual one-class-against-all values, in a weighted or unweighted fashion, for metrics that are implemented for binary classification only?