Q-Learning for MDPs with General Spaces: Convergence and Near Optimality via Quantization under Weak Continuity