[codex] Add structured Fama/French data provider

[dachent]

Summary

• add a structured Kenneth French Data Library provider through scripts/fetch_famafrench.py
• support direct current _CSV.zip feeds and direct archive _CSV.zip URLs
• update aMDT docs and skill metadata to include Fama/French routing and normalization guidance

Why

Fama/French factors and portfolio data are useful for research and backtesting, but aMDT is designed to stay on public structured feeds instead of HTML scraping. This change adds direct support while keeping the provider runtime strictly limited to structured zip feeds.

What changed

• add scripts/fetch_famafrench.py
• add references/famafrench.md
• update README.md, SKILL.md, agents/openai.yaml, and references/normalization.md
• reject _TXT.zip inputs and Kenneth French HTML pages at runtime

Validation

• python -m py_compile across scripts/*.py
• live current-feed tests:
  • F-F_Research_Data_Factors_CSV
  • F-F_Research_Data_Factors_daily_CSV
  • 6_Portfolios_2x3_CSV
• live archive test:
  • direct 2025 Fama/French 3 Factors archive _CSV.zip URL
• guardrail tests:
  • _TXT.zip fails with a structured-feed-only error
  • data_library.html fails with a direct-zip-required error
• confirmed no runtime code path fetches Kenneth French HTML pages