feat: ROS2 point reordering

[contagon]

Handle reordering PointCloud2 types a lot more robustly.

Previously, if the point cloud was column major but not ring sequential, FORM would likely crash. Now FORM handles this by either

1. Adding missing points to the end of the scanline
2. Using a predefined lidar model formats to properly handle the nonsequential points.

[Copilot]

Pull request overview

This PR improves robustness of ROS 2 PointCloud2 ingestion for FORM by introducing a two-step pipeline (parse → infer format → reorder) and supporting predefined LiDAR model formats to handle non-sequential column-major ring ordering without crashing.

Changes:

• Added RawPoint parsing (load_pc2) and new inference/reordering utilities (infer_lidar_size, infer_lidar_order, reorder) for more resilient organized-scan construction.
• Introduced LidarFormat + a small registry of predefined models (LIDAR_FORMATS) and a new lidar_model ROS parameter.
• Updated launch config, docs, and dev task to use lidar_model and allow “infer/auto” defaults for geometry/range parameters.

Reviewed changes

Copilot reviewed 7 out of 7 changed files in this pull request and generated 10 comments.

Show a summary per file

File	Description
ros/src/ros_pc2.h	Adds RawPoint parsing + inference + reordering logic for robust PointCloud2 handling
ros/src/node.hpp	Stores optional inferred/user LiDAR format on the node
ros/src/node.cpp	Adds lidar_model parameter and runtime inference of LiDAR sizing/ordering before registering scans
ros/src/format.hpp	Adds LidarFormat, helper mapping utilities, and predefined model registry
ros/launch/odometry.launch.py	Adds lidar_model and switches geometry/range defaults to auto/infer sentinels
ros/pixi.toml	Updates example task to use lidar_model
ros/README.md	Updates usage/docs around inference and model-based configuration

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.