feat(providers/llm): real OpenAI/Codex adapter (LLM_FALLBACK default)

[wesleysimplicio]

Context

CodexProvider at lib/providers/llm.ts:48-52 is a mock. Per PROVIDERS.md, OpenAI/Codex is the standard fallback for every Claude-defaulted task. The .env.example has LLM_FALLBACK=codex.

Goal

Real OpenAI adapter wrapping the chat completions API.

Acceptance criteria

• Adapter uses openai SDK
• Reads OPENAI_API_KEY from env
• Defaults to gpt-5.1-mini (cheap fallback) for caption/translation, gpt-5.1 for script/compliance
• Real tokens + cost_usd + latency_ms
• Retry + timeout policy
• Skipped when DRY_RUN=true
• E2E spec with mocked HTTP asserts request shape

Files

• lib/providers/llm.ts:48-52
• package.json (new dep openai)

[wesleysimplicio]

Closed after PR #40 follow-up stabilization in commit f40b58d. Current main passes npm run typecheck and npm run test:e2e (119 Playwright specs green), and this tracker item is being retired per repository cleanup.