Investigating Bayesian Neural Network Dynamics Models for Model-Based Reinforcement Learning