- geometry/path.py: full type annotations for Path class
- geometry/layer.py: type hints for Layer, VerticalLayer, VerticalLayersManager
- geometry/print_point.py: full type annotations for PrintPoint
- slicers/base_slicer.py: type hints for BaseSlicer base class

Uses Python 3.9+ style hints with `from __future__ import annotations`
for forward references and modern union syntax (X | Y).

- convert % formatting to f-strings
- fix SIM102/SIM103 (collapsible if)
- fix B007 (unused loop vars)
- fix B904 (raise from err)
- update mesh_bounding_box -> bounding_box(vertices)
- resolve circular imports via direct module imports

- mesh.edge_faces() now takes edge tuple instead of u/v kwargs
- fix Path import in layer.py (needed at runtime for isinstance)
- fix seams_align import in generate_brim.py

Major refactoring to use typed dataclasses instead of unstructured dicts.
All dataclasses inherit from compas.data.Data for COMPAS serialization.

New dataclasses:
- GcodeParameters: typed G-code params with 25+ fields and defaults
- PrintPath: wraps list[PrintPoint] with iteration/indexing
- PrintLayer: wraps list[PrintPath] with iteration/indexing
- PrintPointsCollection: replaces dict[str, dict[str, list[PrintPoint]]]

Converted to dataclasses:
- PrintPoint: @DataClass(Data) with __data__/__from_data__
- Path: @DataClass(Data) with legacy dict format compat
- Layer/VerticalLayer: @DataClass(Data) with inheritance

Key changes:
- Integer indices instead of string keys ('layer_0' -> [0])
- Access: collection[0][1][2] vs dict['layer_0']['path_1'][2]
- Type-safe with IDE autocomplete
- printpoints_dict property for backward compat
- Updated all print organizers and utilities

…ad/cut

- new config.py: SlicerConfig, InterpolationConfig, GcodeConfig, PrintConfig dataclasses
- enums: GeodesicsMethod, UnionMethod for type-safe config
- full type hints: planar/interpolation/scalar_field/uv slicers
- full type hints: planar/interpolation/scalar_field print organizers
- fix direct imports in planar_slicer (create_planar_paths)
- fix mypy issues (optional types, null checks)

- type hints: get_param, defaults_layers, defaults_gcode, defaults_interpolation_slicing
- add min_layer_height, max_layer_height to layers defaults
- add filament_diameter key (underscore) alongside legacy space key
- deprecation note on GcodeParameters (prefer GcodeConfig)

- add _numpy_ops.py with vectorized helpers (KDTree, batch distances)
- add test_performance.py with benchmarks + regression tests
- vectorize gradient.py (vertex/face/edge gradient, divergence)
- vectorize topological_sorting.py (point cloud neighbor check)
- add KDTree to mesh_attributes_handling.py (restore_mesh_attributes)
- pre-factor sparse matrix in geodesics.py heat diffusion (250x speedup)
- vectorize data_smoothing.py iterative loop
- vectorize sort_paths_minimum_travel_time.py (seam adjustment)
- vectorize seams_align.py (distance calculations)
- add numba, pytest-benchmark to deps

Test fixtures use Sphere (Box ignores u,v params):
- small: ~180 faces
- medium: ~2k faces
- large: ~8k faces

Benchmark results:
- vertex gradient (8k faces): ~860μs
- divergence (2k faces): ~125μs
- batch closest points (1k pts): ~650μs

convert list back to Vector in set_ppt_up_vec

curved slicing: 22s -> 13.5s (38% speedup)

curved slicing: 13.5s -> 12s

- generate_brim: CGAL offset_polygon w/ pyclipper fallback
- offset_polygon_with_holes for complex slices (CGAL-only)
- remesh_mesh via CGAL trimesh_remesh for mesh quality

- add get_cgal_HEAT_geodesic_distances using compas_cgal.HeatGeodesicSolver
- add HEAT_CGAL enum to GeodesicsMethod
- CGAL heat ~2.1s vs libigl ~3.1s in full workflow (30% faster)