[IDEA]: apply FlyWeight design pattern to transformation matrices

[ktro2828]

Description

Currently, FrameGroundTruth keeps transformation matrices including sensor related, but it is static that coordinates of sensors with respect to base link. Therefore it is useless to keep those in every frame.

• Current implementation

class FrameGroundTruth:
    ...
    transforms: TransformsDict ...including ego->map, ego->lidar, ego->camearaN,...

• Proposal

class SceneGroundTruth:
    data_root: str
    frames: list[FrameGroundTruth]
    sensors: dict[FrameID, SensorCalibration> ...including sensor information
    
# ----------------------------------------------------------
class FrameGroundTruth:
    ...
    ego2map: HomogeneousMatrix | None

class SensorCalibration:
    frame_id: FrameID
    modality: SensorModality
    position: ArrayLike; 3
    rotation: Quaternion
    camera_intrinsic: ArrayLike; 4 | None
    camera_distortion: ArrayLike; 4 | None

Purpose

Possible approaches

Definition of done