feat(detector): implement AI data exfiltration detector (Epic 4-F3)

[KbWen]

Implements the AI Data Exfiltration Detector (Epic 4-F3) to audit code for LLM prompt exfiltration, MCP tool leaks, and public web writes.

[KbWen]

🔒 Security Hardening & Evasion Defenses Applied (加固與規避防禦已完成)

This PR has been successfully hardened against all bypass vectors identified by security experts and Tenth Man auditors:

🛠️ Key Improvements (關鍵改進項目):

1. SSRF IP Normalization (SSRF IP 標準化): Normalized IPv4 formats (decimal, octal, hexadecimal, and IPv6-mapped Hex e.g., ::ffff:a9fe:a9fe -> 169.254.169.254) to prevent cloud metadata SSRF evasion.
2. Case Normalization (大小寫正規化): Enforced lowercase conversion at path and hostname checks to prevent case-mixing protocol bypasses (e.g. HTTP://).
3. TOCTOU Defenses (時間差防範): Cleared variable taint validation status instantly upon any assignment, modification, or loop iteration to prevent variable re-assignment reuse bypasses.
4. Dummy Validation Block (防範虛假驗證): Restricted validation flags to standalone expression checks and control flow conditions (e.g. if (validate_path(x))), blocking ignored boolean validation return values.
5. JS AST Visitor Hardening (JS 走訪器加固): Added support for JS template literal interpolation, object destructuring in MCP parameters, and monitored file transfer utilities (fs.copyFileSync, fs.renameSync).

🧪 Verification (驗證結果):

• All Tests Pass: 281/281 unit tests passed successfully on Windows env.
• Self-Scan: Local self-scan achieved Grade A (100/100) with 0 warning findings.

Ready for final review and merge!