Reinforcement learning-guided optimization of critical current in high-temperature superconductors