Add face keypoints/animal pose ops & Extend ops for frame-sequence input

[Qirui-jiao]

Add ops:

• video_animal_pose_mapper: Detect quadruped animal pose on the video with ViTPose.
• video_face_keypoints_mapper: Detect face keypoints (98 points) on the video with LDEQ_RwR.

Enable existing ops to support frame-sequence input:

• video_camera_calibration_static_moge_mapper
• video_depth_estimation_mapper
• video_hand_reconstruction_hawor_mapper
• video_whole_body_pose_estimation_mapper

[gemini-code-assist]

Code Review

This pull request introduces two new video mappers, VideoAnimalPoseMapper and VideoFaceKeypointsMapper, for detecting animal poses and face keypoints, respectively. It also updates several existing video mappers to support input from pre-extracted frames. The review feedback identifies several critical issues, including an AttributeError in the animal pose mapper, a missing comma in a category list, and incorrect return types in multiple mappers that would break the processing pipeline. Additionally, there are suggestions to simplify video name extraction and improve the robustness of path handling.