Individual Fairness in Bayesian Neural Networks