Camera Module Refactor, added image rectification

[alexlin2]

Camera Calibration and Image Rectification Refactor

Summary

Refactored the camera calibration pipeline to centralize camera info publishing and add
optional image rectification support.

Changes

🎯 Centralized Camera Info Publishing

• Moved camera_info and camera_pose publishing from DepthModule to
  ConnectionModule
• ConnectionModule now handles all camera metadata publishing, improving separation of
  concerns
• Camera info is loaded once at initialization and reused for every frame

🔧 Added Image Rectification

• Implemented optional image undistortion using OpenCV's cv2.undistort
• Configurable via rectify_image flag in ConnectionModule (default: True)
• When enabled, distortion coefficients are zeroed in published camera_info since images
  are already rectified

📦 New Utility Functions

• load_camera_info(): Loads ROS-style YAML → LCM CameraInfo message
• load_camera_info_opencv(): Loads ROS-style YAML → OpenCV K matrix and distortion
  coefficients
• rectify_image(): Applies undistortion to Image objects while preserving metadata

🔄 Dynamic Camera Configuration

• DepthModule now subscribes to /go2/camera_info instead of requiring manual intrinsics
• Metric3D automatically initializes when camera_info is received
• No more hardcoded camera parameters in module initialization

Benefits

• ✅ Cleaner architecture: Each module has a single, well-defined responsibility
• ✅ Plug-and-play: DepthModule works without manual configuration
• ✅ Better accuracy: Optional rectification improves computer vision algorithms
• ✅ Consistent calibration: All perception modules use the same camera parameters
• ✅ Easier maintenance: Camera parameters loaded from single YAML source

[leshy]

looks good!