Text this: Combining Prior Knowledge and Reinforcement Learning for Parallel Telescopic-Legged Bipedal Robot Walking