Text this: Semantic Fusion and Depth‐to‐Volume Backend Optimization for Robotic Simultaneous Localization and Mapping