Model-enhanced Contrastive Reinforcement Learning for Sequential Recommendation