A Framework for Designing Reward Functions: From Objectives to Features to Human-Aligned Reward Functions