tfb_batch_normalization function

ComputesY = g(X) s.t. X = g^-1(Y) = (Y - mean(Y)) / std(Y)