traversability_generator3d: A C++ library for 3D traversability estimation from MLS maps

Autonomous navigation in unstructured outdoor environments requires robot-specific reasoning about where motion is physically feasible. While 2.5D elevation maps lose information in multi-layered terrain and voxel/octree approaches can be memory-intensive without explicit traversability semantics, Multi-Level Surface (MLS) maps provide a compact representation that preserves multiple vertical surface hypotheses per cell. However, MLS encodes structure rather than actionability. traversability_generator3d builds upon MLS by transforming geometric surface information into robot-aware 3D traversability maps through plane fitting, slope and step-height analysis, footprint-aware collision checks, orientation constraints, and optional soil semantics. The resulting TraversabilityMap3d enables planners to reason about safe motion, uncertainty, and exploration boundaries in complex environments such as rubble fields, vegetation, construction sites, and planetary terrain, while maintaining a modular separation between terrain representation and traversability reasoning