Reinforcement Learning Meets Large Language Models: A Survey of Advancements and Applications Across the LLM Lifecycle