Reinforcement learning

Basic implementation of methods in Sutton and Barto's Reinforcement Learning book.

We consider numba for gridworld experiments and Jax where gradients are required.