How to Enable Uncertainty Estimation in Proximal Policy Optimization