Added a new python script to inference with webcam

[lohithkumaar]

feat: add webcam inference demo script

This commit introduces, a script that enables real-time inference using the VILA model directly from a webcam feed.

Key features:

• Captures frames from the default webcam using OpenCV.
• Handles image preprocessing (temporary file saving).
• Runs inference using the loaded VILA model.
• Supports --run-loop for continuous capture and analysis.

This simplifies testing and demonstration on edge devices like the Jetson AGX Orin.