Score-Based Diffusion Policy Compatible with Reinforcement Learning via Optimal Transport