fix(llm): do not route custom OpenAI apiBase to responses API unless explicitly configured

[Alphaxiaoteng]

Description

Fixes #12995

Problem

When configuring custom OpenAI-compatible proxies / endpoints (apiBase) with model names starting with gpt-5 or o-series (e.g. o3-mini, o1), Continue incorrectly attempted to use the OpenAI Responses API (/responses) instead of Chat Completions (/chat/completions). Most third-party OpenAI-compatible endpoints only implement /chat/completions, causing requests to fail with 404 or unsupported endpoint errors.

Solution

1. In core/llm/index.ts (canUseOpenAIResponses), only auto-enable Responses API when apiBase is the official OpenAI API (or empty/default), while respecting explicit overrides (useResponsesApi: true forces Responses API, useResponsesApi: false forces Chat Completions).
2. In packages/openai-adapters/src/apis/OpenAI.ts (shouldUseResponsesEndpoint), normalize apiBase check to official OpenAI API endpoint and honor explicit useResponsesApi: true override on custom endpoints.
3. Added unit test suites verifying routing behavior for official vs custom apiBase and explicit useResponsesApi configuration options.

Checklist

• Tested locally and added automated tests
• Conforms to project conventions and formatting

[github-actions]

Thank you for your submission, we really appreciate it. Like many open-source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution. You can sign the CLA by just posting a Pull Request Comment same as the below format.

---

I have read the CLA Document and I hereby sign the CLA

---

_{You can retrigger this bot by commenting recheck in this Pull Request. Posted by the CLA Assistant Lite bot.}