Lidar detection and classification integration

[guillaume-byte]

Integrate Lidar detection and classification into the core WeightsLab library for the next release.

Scope:

• Add a Lidar detection and classification pipeline (preprocessing, detection, classification) to the codebase.
• Provide pretrained models or training scripts and clear instructions for retraining.
• Expose a stable inference API (Python and REST) for downstream consumers.
• Add unit/integration tests, performance benchmarks, and CI coverage.
• Update documentation and release notes.

Acceptance criteria:

• Detection and classification can be run via the public API with example scripts.
• Tests covering key pipeline stages are added and pass on CI.
• Documentation includes usage examples and performance notes.

Planned tasks:

• Data ingestion and preprocessing for Lidar inputs
• Implement detection and classification modules and integration into pipeline
• Add model training scripts and provide at least one pretrained model
• Implement inference API and example scripts
• Add tests, benchmarks, and CI workflows
• Update docs and changelog for next release