A Shared Low-Rank Adaptation Approach to Personalized RLHF