fix: resolve ZeroDivisionError in mvp.py and enable dynamic video loading

[Kunal-Somani]

Resolves #13

Objective

Resolves the ZeroDivisionError in mvp.py identified in #13 and allows for dynamic video path loading.

Technical Changes

1. Updated cv2.VideoCapture() in main() to utilize the passed video_path argument instead of a hardcoded fv.mp4 string.
2. Implemented an early-exit guard clause in getPercentages() to safely return an empty dictionary if the predictions array is empty. This prevents a fatal crash when processing edge cases (e.g., corrupted files or frames with no detectable faces).

Validation

Verified locally by executing mvp.py against a nonexistent/empty video file. The application successfully logs No predictions generated. Returning empty percentages. and exits cleanly instead of throwing a Python traceback.

Terminal Proof:

(venv) kunal@kunal-OMEN-by-HP-Gaming-Laptop-16-wf1xxx:~/.../facial-sentiment-analysis-api$ python3 mvp.py dummy_nonexistent_video.mp4
# ... [TensorFlow initialization logs omitted for brevity] ...
[]
No predictions generated. Returning empty percentages.