Robustness of Generalized Learning Vector Quantization Models against Adversarial Attacks