docs: clarify OpenAI-compatible endpoint configuration - #321

What changed

Add an OpenAI-compatible endpoint example to the CLI guide. It explains why a remote server needs --backend openai, how to choose the API root for --base-url, and why the full /chat/completions URL is wrong. It also covers the server's model ID, the unset-key fallback, and the evaluation sandbox.

Provider-specific URLs are unchanged. In particular, /v1 is not required by every provider.

Fixes #315.

Validation

• Ran the real OpenAIChatDecoder against a local HTTP server with the OpenAI client. All four cases passed: roots with and without /v1, each with a configured key and an unset key. The captured requests used the expected chat path, model ID, and authentication fallback.
• Repository pre-commit checks passed for docs/cli.md; Python and YAML hooks correctly skipped this Markdown-only change.
• git diff --check passed.
• No paid model calls were made. The full repository test suite was not run for this docs-only change.

AI assistance

Prepared and checked with AI assistance. No independent human code review is claimed.