Add confidence score explanation to output metadata

[dreamlessx]

What

When the pipeline outputs a confidence score, include a human-readable explanation of what contributes to it (identity similarity, landmark accuracy, face detection confidence).

Why

Users see a confidence number (e.g., 0.82) but don't know what it means or how to improve it. Breaking it down into components helps clinicians understand and trust the output.

Suggested approach

1. Extend the result dict from pipeline.generate() to include a confidence_breakdown field:

   {
       "confidence": 0.82,
       "confidence_breakdown": {
           "face_detection": 0.95,
           "identity_preservation": 0.78,
           "landmark_accuracy": 0.88,
           "mask_coverage": 0.91,
       }
   }

2. Add a --explain flag to the CLI that prints the breakdown
3. Show the breakdown in the Gradio output panel

Difficulty: 🟡 Intermediate