feat: add support for Codex

Description

Codex is a popular agentic platform (https://chatgpt.com/codex). Tier-2 target under epic #51 — bring VGV best practices to more platforms and users.

Codex is a "universal" agent: it reads .agents/skills/ directly, so once the repo is published to the open standard (#50), skills land in Codex for free. The remaining work is MCP and hooks.

Scope

• Skills — verified to load from .agents/skills/ and to degrade where a referenced MCP tool is absent (Phase 0).
• MCP — wire dart and very-good-cli servers via Codex ~/.codex/config.toml.
• Hooks — port the enforcement layer (auto analyze/format, CLI gating) to Codex hooks. Codex loads hooks from hooks.json or inline [hooks] in config.toml, with PreToolUse / PostToolUse / UserPromptSubmit events. Caveats: experimental, feature-flagged ([features] codex_hooks = true), disabled by default, not available on Windows. Adapt ${CLAUDE_PLUGIN_ROOT}-style path vars as needed.
• Subagent — Codex has full subagent support (GA March 2026, custom agents as TOML in ~/.codex/agents/). Convert the flutter-reviewer agent to Codex TOML so it dispatches natively.
• CI — loader job that installs the skills and asserts they load in Codex.

Docs: https://developers.openai.com/codex/hooks and https://developers.openai.com/codex/subagents

Requirements

• Skills load in Codex via .agents/skills/
• dart + very-good-cli MCP wired via config.toml
• Enforcement hooks ported to Codex hook format (with Windows / feature-flag caveats documented)
• Codex loader CI job added
• All CI/CD checks are passing.
• There is no drop in the test coverage percentage.

Additional Context

Part of epic #51. Tier 2 / wave 1 alongside Gemini CLI (#125) and OpenCode (#127).