Fixes for deployment, around prod server in gunicorn

[KB-perByte]

Fixes -

1. Fix hardcoded paths: Replace user-specific paths with ~/AI-Issue-Triage default
2. Fix webhook security bypass: Reject webhooks when GITHUB_WEBHOOK_SECRET is missing instead of accepting
3. Fix unbounded issue fetching: Limit duplicate check to 200 issues max and 90-day window to prevent rate limiting
4. Fix destructive label removal: Only remove bot-managed labels (Type :, Severity :, etc.), preserve user-added labels
5. Fix inefficient double-clone: Use local repo for repomix instead of --remote to avoid cloning twice
6. Fix disk space leak: Add robust error handling and fallback cleanup for temp directories
7. Add production server warning: Warn users that Flask dev server is not production-ready, recommend gunicorn/uvicorn
8. Fix silent error masking: Re-raise exceptions in webhook handlers so GitHub retries on failure (500 vs 200)
9. Fix cleanup edge cases: Wrap all os.unlink() and shutil.rmtree() calls in try/except blocks

[shvenkat-rh]

@KB-perByte can we merge PR #3 before this?

[KB-perByte]

Your call @shvenkat-rh i don't mind.

[shvenkat-rh]

Please review PR #3