ModelMirror v0.1.0

Initial public release candidate for ModelMirror.

Highlights:

• Offline prompt-pack debugger for testing agent expectations before LLM runtime.
• Local Streamlit harness with no telemetry, no default network calls, and no real tool execution.
• Doctor, run, and diff commands for repeatable prompt-pack review.
• CI smoke checks for Python 3.10, 3.11, and 3.12.
• Agent and framework integration notes for Codex-style workflows.

This release is intentionally small: it makes expectations visible before we later expand ModelMirror into a fuller agent runtime testing harness.