Enhance error metrics of binary classification

[PetrToman]

Encog already supports a parameter search for SVM utilizing a simple error metric:

if( actual==ideal ) correct++;   // EncogUtility.calculateClassificationError()

However, there is a number of other evaluation functions (like F1-score or BAC), each of which is useful in a different scenario - see: http://www.csie.ntu.edu.tw/~cjlin/libsvmtools/eval/index.html

This enhancement would be useful epsecially for cross validation - after regularization is implemented (issue #28).