MVAUtils: Adding support for XGBoost

Adding derived class to support XGboost node/forest in MVAUtils. A python utility to convert the model saved by XGBoost training to our TTree format is also provided. Implementation follows closely with what was done in lgbm, presently only support regression and binary classification.

Reconstruction/MVAUtils/MVAUtils/BDT.h

 /*
-  Copyright (C) 2002-2019 CERN for the benefit of the ATLAS collaboration
+  Copyright (C) 2002-2020 CERN for the benefit of the ATLAS collaboration
 */
 
 #ifndef MVAUtils_BDT_H
@@ -22,7 +22,7 @@ namespace MVAUtils
 {
   class IForest;
 
-   /** Simplified Boosted Regression Tree, support TMVA and lgbm.
+   /** Simplified Boosted Regression Tree, support TMVA, lgbm, and xgboost.
    * Holds a forest (vector of top nodes of each decision tree) and a
    * constant offset or set of weights (not always used).
    *
@@ -40,6 +40,7 @@ namespace MVAUtils
   public:
     /** Constructor. The input tree must be created with
      *  convertLGBMToRootTree.py (for lgbm training) or with
+     *  convertXGBoostToRootTree.py (for xgboost training) or with
      *  convertXmlToRootTree (for tmva training)
      **/
     explicit BDT(TTree *tree);//ctor TTree