fix: match gpt-5 and o-series model names case-insensitively

[Igralino]

Summary

• isOSeriesOrGpt5PlusModel in core/llm/llms/OpenAI.ts and the equivalent check in packages/openai-adapters/src/apis/OpenAI.ts are case-sensitive, so model names like GPT-5, Gpt-5-mini, or O1 slip past the regex and never get their max_tokens rewritten to max_completion_tokens.
• In practice this breaks Azure OpenAI deployments, which are commonly named in uppercase (e.g. an Azure deployment literally called GPT-5). The provider then returns 400 Unsupported parameter: 'max_tokens' is not supported with this model. Use 'max_completion_tokens' instead.
• Fix: add the i flag to the regexes in isOSeriesOrGpt5PlusModel, and normalize to lower-case before the startsWith("o") / includes("gpt-5") / startsWith("o3") checks in the adapter. Added uppercase/mixed-case test cases.

This mirrors the existing case-insensitive pattern in packages/openai-adapters/src/apis/openaiResponses.ts (/^(?:gpt-5|gpt-5-codex|o[0-9])/i).

---

Summary by cubic

Make GPT-5 and o‑series model detection case-insensitive. This fixes max_tokens → max_completion_tokens rewriting for uppercase Azure deployments (e.g., GPT-5) and prevents 400 errors.

• Bug Fixes
  • core/llm/llms/OpenAI.ts: add i flag to o[0-9]+ and gpt-[5-9] regexes.
  • packages/openai-adapters/src/apis/OpenAI.ts: use model.toLowerCase() for startsWith("o"), includes("gpt-5"), and startsWith("o3").
  • Extend unit tests with uppercase/mixed-case models (GPT-5, Gpt-5.4, O1, O3-mini).

^{Written for commit 0576ec7. Summary will update on new commits.}

[github-actions]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[cubic-dev-ai]

No issues found across 3 files

[Igralino]

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