fix(core): emit message before function_calls in toResponsesInput

[YgorLeal]

Summary

Fixes #11994 — Resolves 400 Bad Request: 'Missing reasoning item' errors when using reasoning models (o1, o3) through the OpenAI Responses API.

Problem

In toResponsesInput, when an assistant message contains both text and tool calls, function_call items were emitted before the text message item. This separates a preceding reasoning item from its associated message, violating the Responses API's sequencing requirements:

// Before (broken): reasoning → function_call → function_call → message
// After (fixed):   reasoning → message → function_call → function_call

Solution

Swapped the emission order in the case "assistant" block so the text message is pushed before the function_call items. This keeps the reasoning item adjacent to its message output.

Test plan

• Verify reasoning models (o1/o3) no longer produce 400 errors during multi-step tool-calling turns
• Verify non-reasoning models still work correctly with tool calls
• Existing test suite passes (lint-staged/prettier verified on commit)

---

Summary by cubic

Resolves #11994 by fixing Responses API sequencing in toResponsesInput: emit the assistant text message before function_call items. This keeps a preceding reasoning item adjacent to its message and prevents 400 "Missing reasoning item" errors with o1/o3.

^{Written for commit f272d7e. 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 1 file

[YgorLeal]

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