Finding Influential Training Samples for Gradient Boosted Decision Trees