Agentic Reinforcement Learning (RL) for Large Language Models (LLM).Markov Decision Processes (MDPs)