Implemented multi-class classification for CUDA and CPU backend.

tmva/tmva/inc/TMVA/DNN/Architectures/Cpu.h

tmva/tmva/inc/TMVA/DNN/Architectures/Cpu/CpuBuffer.h

tmva/tmva/inc/TMVA/DNN/Architectures/Cuda.h

tmva/tmva/inc/TMVA/DNN/Architectures/Cuda/CudaBuffers.h

tmva/tmva/inc/TMVA/DNN/Architectures/Cuda/Device.h

tmva/tmva/inc/TMVA/DNN/Architectures/Reference.h

tmva/tmva/inc/TMVA/DNN/Functions.h

tmva/tmva/inc/TMVA/NeuralNet.icc

tmva/tmva/src/DNN/Architectures/Cpu/CpuBuffer.cxx

tmva/tmva/src/DNN/Architectures/Cpu/LossFunctions.cxx

tmva/tmva/src/DNN/Architectures/Cpu/OutputFunctions.cxx

tmva/tmva/src/DNN/Architectures/Cuda/ActivationFunctions.cu

tmva/tmva/src/DNN/Architectures/Cuda/Arithmetic.cu

tmva/tmva/src/DNN/Architectures/Cuda/CudaBuffers.cxx

tmva/tmva/src/DNN/Architectures/Cuda/CudaMatrix.cu

tmva/tmva/src/DNN/Architectures/Cuda/Dropout.cu

tmva/tmva/src/DNN/Architectures/Cuda/Kernels.cuh

tmva/tmva/src/DNN/Architectures/Cuda/LossFunctions.cu

tmva/tmva/src/DNN/Architectures/Cuda/OutputFunctions.cu

tmva/tmva/src/DNN/Architectures/Cuda/Propagation.cu