feat: add prompt engineering comparison demo

[rlippmann]

What changed

• Added scored Demo 07: demos/07_llm_prompt_engineering_comparison.py
• Registered demo 07 in demos/run_demo.py as an evaluative demo
• Updated demos/README.md demo overview and evaluative enumeration to include demo 07
• Added focused tests for demo 07 prompt construction and report semantics in tests/test_07_llm_prompt_engineering_comparison.py
• Added mocked-LLM integration coverage for demo 07 in tests/test_llm_demos.py

Why this was needed

This adds a three-way prompt-engineering comparison demo to show that context-compiler complements prompt engineering: prompting quality matters, and authoritative compiled state adds additional reliability under transcript drift and contradiction pressure.