Port LiDAR Motion Distortion Correction to Awkernel #760

Implement LiDAR motion distortion correction in Awkernel using per-point timestamps and vehicle motion estimates. Transform points acquired at different times into a common scan reference frame, matching the behavior of the selected Autoware reference implementation and correction mode.

Scope:

• Interpret per-point timestamps, including their units and relationship to the scan timestamp
• Buffer timestamped vehicle twist estimates and IMU angular velocity when required by the selected correction mode
• Estimate relative motion between each point’s acquisition time and the scan reference time
• Apply sensor extrinsics and motion compensation consistently with the defined coordinate-frame conventions
• Preserve required point fields and document output timestamp semantics
• Define behavior for insufficient motion-history coverage, invalid timestamps, and excessive interpolation or extrapolation gaps

Acceptance criteria:

• Corrected point coordinates match the reference implementation within predefined tolerances under identical inputs and configuration.
• Tests cover stationary conditions, constant translation, constant rotation, and combined motion.
• Synthetic tests with known trajectories verify correction direction, reference time, and sensor extrinsic handling.
• The stationary case leaves point coordinates unchanged within numerical tolerance.
• Tests cover non-identity sensor extrinsics, scan-boundary timestamps, and missing or stale motion estimates.
• Unsupported or insufficient motion data triggers a documented rejection or explicitly flagged fallback rather than silent correction.
• Execution time and memory usage are measured across representative point counts and motion-input rates.
• Test datasets, parameters, and reproduction steps are documented.

Out of scope:

Sensor calibration estimation, IMU bias estimation, changes to gyro odometry or EKF algorithms, and full-pipeline accuracy evaluation.