Optimizing Protein Design with Reinforcement Learning-Enhanced pLMs: Introducing DPO_pLM for Efficient and Targeted Sequence Generation