Robust Explainability: A Tutorial on Gradient-Based Attribution Methods for Deep Neural Networks