Tiny Reinforcement Learning for Quadruped Locomotion using Decision Transformers