feat: add copilot fetch adapter

[NateSmyth]

What does this PR do?

• Adds a fetch adapter for Copilot chat models to handle chat API's reasoning format and translate it for AI SDK.
• Works with Claude and Gemini.
• Passing through thinking_budget and the interleaved thinking header for claude.
• Works with the existing openai-compatible sdk

How did you verify your code works?

• Testing multi-turn thinking with both claudes and gemini.
• Tested cycling between models to ensure no cross contamination (copilot/copilot and copilot/noncopilot).
• Logged requests/responses to confirm headers, provider options, metadata, and content were all correct.
• Tested gpt-5 on /responses to make sure I didn't accidentally break this.

---

Would fix#6864
also fixes#8417

[github-actions]

Thanks for your contribution!

This PR doesn't have a linked issue. All PRs must reference an existing issue.

Please:

1. Open an issue describing the bug/feature (if one doesn't exist)
2. Add Fixes #<number> or Closes #<number> to this PR description

See CONTRIBUTING.md for details.

[github-actions]

The following comment was made by an LLM, it may be inaccurate:

Based on my search, I found potentially related PRs that address similar copilot and reasoning functionality:

1. feat(opencode): add copilot specific provider to properly handle copilot reasoning tokens (PR feat(opencode): add copilot specific provider to properly handle copilot reasoning tokens #8900)

   • Directly related to handling copilot reasoning tokens, which overlaps with the new PR's focus on handling reasoning format

2. Feature/OpenAI compatible reasoning (PR Feature/OpenAI compatible reasoning #5531)

   • Related to reasoning format handling for OpenAI-compatible APIs, which is relevant since the new PR works with the openai-compatible SDK

3. fix(provider): use snake_case for thinking param with OpenAI-compatible APIs (PR fix(provider): use snake_case for thinking param with OpenAI-compatible APIs #10109)

   • Related to the thinking_budget parameter handling mentioned in the new PR

4. fix: Prevent excessive Copilot premium request consumption (PR fix: Prevent excessive Copilot premium request consumption #8721)

   • Related to copilot-specific handling

These PRs may be addressing the same or overlapping concerns around copilot reasoning support. You may want to review PR #8900 in particular as it seems most closely related to this new adapter work.

[caozhiyuan]

@NateSmyth this be merged #8900. But there are some small problems.

[NateSmyth]

@caozhiyuan What do you mean? Problems with this PR?
Or with the merged PR #8900?

[caozhiyuan]

Similar function PRs have been merged. @NateSmyth

[caozhiyuan]

But yours seems simpler; there's still a problem with that PR.

[github-actions]

Closing this pull request because it has had no updates for more than 60 days. If you plan to continue working on it, feel free to reopen or open a new PR.