OpenRubrics: Towards Scalable Synthetic Rubric Generation for Reward Modeling and LLM Alignment