Add content tags, content type filtering, docs overhaul, and config improvements

[ehsaniara]

Summary

• Add deny_content_tags policy layer, developers mark files with // NO_LLM
  to hard-block them from LLMs (no prompt, no override)
• Add skip_content_types to skip scanning binary/non-text payloads
  (images, PDFs, archives), avoids false positives since Egressor has no OCR
• Remove auto-created default config, if no config is found, Egressor prints
  where it looked and how to set one up, then exits
• Rewrite all documentation for clarity:
  • config.yaml , inline comments on every field explaining what it does
  • README.md , reorganized with four blocking layers explained, Windows
    step-by-step, content tags with code examples
  • docs/ , split design.md into four focused docs: design overview, policy
    layers, data flow scenarios, and component details