TEST: Add tests of several classifiers for the OpenCV framework in the TrainImagesClassifier application

ENH: Handling ParameterType_Double SVM parameters for the OpenCV framework in the TrainImagesClassifier application

BUG: Removing remaining static declarations. The static mutex has been moved to a non-template base class, while the firstTime static bool has been replaced by unregistering each factory before registering again.

BUG: Removing static declarations in template clas to fix bug #418. Two static declaration remains (not easy to remove) so the bug is not fixed yet

BUG: Neural Network training needs a matrix of labels with one column for each class and not a single column matrix with the class label.

ENH: Remove regression flag in Boosting machine learning model. This model can only be used for classification.

ENH: Expose K parameter to control number of neighbors used in KNN machine learning. Also expose IsRegression parameter.

ENH: Add a parameter IsRegression in BoostMachineLearningModel to indicate if it is a regression model or a classification model.

ENH: Add an ImageClassifier application that classifies an image using a model obtained with TrainMachineLearningImagesClassifier application.

This classes have been gfork and static variables have been removed to avoid dl.close error in applications.
I used itykv4 version of classes to do the fork.
Forked classes are only use in the BinaryMorphological applications. It is still the internal itkv3
which is used in OTB.
Need to report this problem to ITK in JIRA, when the problem is solve on the ITK side we would safely remove these classes and use ITK classes