Improve 3D object detection evaluation

[RaphvK]

This pull request introduces support for optional ground truth inputs in the evaluation pipeline, making the evaluation node more flexible when working with datasets that may not publish all metadata or auxiliary information. The documentation and configuration have also been updated to reflect a more general evaluation naming scheme and improved clarity about supported datasets and tasks.

Key changes:

• Updated the evaluation naming from nuscenes_lidar_object_detection to a more general object_detection_3d throughout the codebase and documentation, reflecting support for multiple datasets and a broader range of object detection tasks. [1] [2] [3] [4] [5] [6]
• Improved documentation to clarify that evaluations can use established metrics across all supported datasets and that optional ground truth is handled gracefully. [1] [2]

Configuration enhancements:

• Added environment variables in .env to select the system under test and dataset, improving configuration flexibility.