Text this: IGRL: Imitation-Guided Reinforcement Learning for Sample-Efficient and Stable Robotic Control