#10 simplify Classifier/Trainer, etc. class hierarchy

[Raul Ramos Pollan]

should be something like

  :::java
  public interface MLLMethod {
     public void fit(List<DataInstances>);   // fits to a set of data instances
     public void score(List<DataInstances>); // scores a set of data instances (assumes it has been fit before)
     public String getModelDump();
     public void setModelFromDump(String dumpedModel);
  }

in classification tasks, mll will compute/apply threshold if needed (recall, for computing AUC is not needed)

when fitting, classification tasks will look for an IdealClass field in DataInstance

when fitting, regression tasks will look for an IdealScore field in DataInstance (must implement)