v0.4.0

Added

• effort parameter for both codex and codex-reply tools (medium, high, xhigh)
• LLM agents auto-select reasoning effort based on task complexity

Changed

• Server version now dynamically reads from package.json instead of hardcoded value
• Installation guide updated with 3-scope prompt and language matching