← Convergence of Q-learning: a simple proof