enrichmcp 0.4.7

Added

• Server-side LLM sampling via ask_llm() with new ask_user() helper.
• prefer_medium_model() for balanced model selection.
• Expanded examples and tests including Redis cache and travel planner.

Changed

• Removed automatic EnrichContext injection; use app.get_context().
• Improved chat agent and memory examples.

Fixed

• Correct context usage in examples and ask_llm helper.