CAF support for TMVA/Keras featrues with root 6-14-04

Tested in root 6-14-04, TMVA supports pyMVA with the following methods.

SKL methods: PyGTB,PyAdaBoost,PyRandomForest

Keras methods: PyKeras (# need to generate Keras model first)

Simple example for training with the above methods attached. TMVAClassification_VBF_BKG_com.C

For us, for application phase of MVA analysis, it's good to have the ability to be able to apply the trained results in CAF. (For training phase, train samples outside root with numpy looks promising already)

Please investigate the possibility to update the root version for CAF and we can see if other actions are needed for application phase. Thank you!

Brgs, Yun-Ju