https://flecart.github.io/notes/tabular-reinforcement-learning/