Text this: Two-Stage Learning of CPG and Postural Reflex Toward Quadruped Locomotion on Uneven Terrain With Simple Reward