Reinforcement

Reinforcement learning algorithms Not well designed yet, but it works. I'll fix it, I promise.

agent.h is for agent going through the maze; SARSAAndQ.cpp provides learning by SARSA and Q-learning algorithms.