Function fpfh_descriptors 

pub fn fpfh_descriptors(
    cloud: &PointCloud,
    feature_radius: f32,
) -> SpatialResult<Vec<FpfhDescriptor>>

Computes an FPFH descriptor for every point in cloud.

The cloud must carry finite unit normals and finite positions. feature_radius is the neighborhood radius used to build each descriptor (≈5× the point spacing is a good start). This is the reusable building block behind FpfhRansacRegistration; compute it once on a keypoint cloud to drive descriptor-based matching cheaply. Independent point descriptors use bounded scoped CPU workers sharing one KD-tree. Small clouds and one-worker staging remain serial; neighbor order, floating-point operation order and output point order are preserved.