https://trunghng.github.io/posts/reinforcement-learning/policy-gradient-theorem/