Research viability of using Dev Proxy with GitHub Models LLMs and handling throttling

We need to research the feasibility of integrating Dev Proxy with LLMs provided by GitHub Models. Specifically, determine if the basic tier of GitHub Models can handle the volume of requests that Dev Proxy generates (such as during API spec generation). Additionally, investigate whether there are any request limits or throttling behaviors that may impact this integration, as our current Dev Proxy implementation did not require throttling logic when used with local models. We should also verify, if our prompts are good enough or if we'd need to think about exposing prompt-sets to make it easier to switch between the models without having to manually replace all prompts in our default files.

If using GH Models turns out to be viable we'll likely need to:

- [ ] Implement handling throttling with proper back-off
- [ ] Add support for auth in the OpenAI language model client

Let's focus on verifying the viability, before we commit to an actual implementation.

ローカルモデル向けの現在の Dev Proxy 統合、API 仕様の生成フロー、OpenAI の言語モデルクライアント、およびデフォルトのプロンプトファイルをレビューしてください。GitHub Models の basic tier が生成されるリクエスト量に対応できるかを判断し、リクエスト制限とスロットリングの動作を文書化し、プロンプトの適合性を評価し、back-off、認証、またはプロンプトセットの変更が必要かどうかを記録してください。

索引モデルが issue の本文から書いたものです。