Feature: Add multi-pass rewriting for higher human scores

[rudra496]

Feature Request

Would be great to support multi-pass rewriting where the text goes through the AI model multiple times to further reduce detection scores.

Implementation Idea

1. First pass: Initial rewrite with anti-detection prompt
2. Second pass: Re-rewrite focusing on burstiness and perplexity
3. Score check: If human score < 90%, do another pass
4. Max 3-5 passes to avoid quality degradation

Additional Context

I implemented something similar in StealthHumanizer (called Ninja mode) and it significantly improved detection bypass rates. Would be happy to collaborate on this.