Towards Piece-by-Piece Explanations for Chess Positions with SHAP