server-fix : accept text tool outputs in responses API

[jesco-absolut]

Overview

This PR fixes:

1. Issue Responses API: function_call_output.output validation throws on valid client input #23557
2. the function for call_output was too strict
3. Only accepted type : "input text"
4. Normalized to chat completions type : "text"

Also added a regression test in tests/test-chat.cpp which covers arrays containing output_text and text. You can run validation run by using uvx cmake --build build --target test-chat

Requirements

• I have read and agree with the contributing guidelines
• AI usage disclosure: : YES - used for local validation assistance; I manually reviewed the final changes.