Neural Additive Models for Location Scale and Shape: A Framework for Interpretable Neural Regression Beyond the Mean