perception: ROS2 Node to convert from 2D to 3D detections

[meshvaD]

Tested some nuscenes rosbags, and tuned clustering parameters based on some videos, but we might need to refine the parameters for our own car's data.

Here is an example birds-eye-view video of the node at work. The colourful points indicate the different objects that are detected. The black bounding outlines are bounding boxes, but the drawn bounding boxes are not cleared after they are redrawn (it seems to be a foxglove problem)

bird_eye.mp4