From Imitation to Interaction: Mastering Game of Schnapsen with Shallow Reinforcement Learning