Conditioning of Reinforcement Learning Agents and its Policy Regularization Application