Feat:Backend Image Processing Blocks the Single-Worker ProcessPoolExecutor

### Describe the feature

Heavy processing tasks like AI tagging + face detection + face clustering are all submitted to this single worker. If a user adds a large folder, every subsequent request (sync, AI tagging enable) is queued behind it.Additionally, in
images.py #L105-L137, images are processed one-by-one in a loop — no batching or parallelism.

### Add ScreenShots

Proposed Fix:
--Increase max_workers or use a proper async task queue (e.g., Celery, RQ, or even asyncio.to_thread)
--Implement batch inference for YOLO/FaceNet instead of per-image model invocations
--Add progress reporting (SSE or WebSocket) to the frontend

### Record

- [x] I agree to follow this project's Code of Conduct
- [x] I want to work on this issue