Reward Scale Robustness for Proximal Policy Optimization via DreamerV3 Tricks