feat: add FastLabel 3D to T4 format

[miursh]

This pull request introduces a new conversion task to the perception_dataset module, adds new label configurations, and includes utility functions to handle rotations. The most important changes include adding a new converter class for FastLabel to T4 data, updating the conversion script to handle the new task, and enhancing the annotation loading and formatting functions.

New Converter Class:

• perception_dataset/fastlabel_to_t4/fastlabel_to_t4_converter.py: Added the FastLabelToT4Converter class to convert FastLabel data to T4 data, including methods to handle directory structures, annotations, and consistency checks.

Conversion Script Update:

• perception_dataset/convert.py: Updated the main function to include a new task "convert_fastlabel_to_t4" that utilizes the FastLabelToT4Converter class.

Annotation Handling Enhancements:

• perception_dataset/fastlabel_to_t4/fastlabel_2d_to_t4_converter.py: Modified the _load_annotation_jsons method to support filtering and loading annotations based on dataset names and delimiters.

Utility Functions for Rotations:

• perception_dataset/utils/transform.py: Added rotation_to_quaternion and quaternion_to_rotation functions to convert between rotation vectors and quaternions.

Label Configuration:

• config/label/object.yaml: Added new label configurations for other_pedestrian and other_vehicle.

[ktro2828]

@miursh Thank you for your PR! Although I left one suggestion, the other parts look good.

[ktro2828]

LGTM