State Encoders in Reinforcement Learning for Recommendation: A Reproducibility Study