KnowRL: Exploring Knowledgeable Reinforcement Learning for Factuality