docs: simplify manual testing spec with design philosophy

Key changes:
- Add "Jupyter Notebooks for CLI Testing" mental model
- Remove HTML comment syntax for review criteria (use plain prose)
- Simplify Phase VI to defer advanced comparison modes
- Emphasize building on existing tools (tryscript + git diff)
- Add design principles: non-interactive, agent-friendly, prose is docs

The core insight: tryscript run --update + git diff already works.
Minimize new features, maximize reuse.

https://claude.ai/code/session_013zTMZFAZESM7uy9oAAxCKN

Author: claude