Remove loading screen implementation and related logic from Marquee

Refactor code structure for improved readability and maintainability

Refactor face detection and tracking in faces.py; optimize performance by reducing redundant detections and using a dictionary for track lookups. Update OpenVINO model usage and remove unused models. Enhance logging for better visibility.

Add export functionality for broadcast as MP4 with progress indication

Add OpenVINO face models, video normalization, and Qwen multimodal model