Document AI output validation and fallback behavior

[MyouzzZ]

Summary

• document the current AI response clean/parse/normalize flow and fallback behavior
• explain malformed output risks for JSON, schema, quiz options, card types, and unsafe text
• list validation layers, safe fallback reason ideas, and backend test cases to add

Closes #13.

Validation

• Reviewed clean_ai_json_response, normalize_cards, fallback cards, and AI generation paths in backend/main.py
• Checked frontend card rendering and quiz option handling in study-copilot-full.html
• git diff --check -- docs/ai-output-validation-fallbacks.md