Explainable AI Security: Exploring Robustness of Graph Neural Networks to Adversarial Attacks