Get composite slices for a video

Retrieves composite slices (shots, scenes, etc.) within a specific video ranked by their similarity to a text query. Given a video_id and text_query, this endpoint performs a drill-down search to find moments within that video. The text query is encoded using the dataset's configured shot-level encoder (e.g., Perception Encoder or other vision-language embedding models), then searched against shot embeddings for that specific video. Returns a ranked list of composite slices ordered by similarity score, including timestamps (start_time_ms, end_time_ms), frame numbers, and representative keyframes.