add reasoning_content to assistant message to consistently extract thinking from response

[pelikhan]

Pioneered by LMStudio

[github-actions]

Comment: The title suggests enhancing the assistant's responses by introducing "reasoning_content" to consistently extract and present the underlying thought process. This could aid in better understanding and potentially improve the quality of responses for users. However, I notice that the body is incomplete or vague, lacking specific details on how this would be implemented. Please provide more information on what "reasoning_content" entails and what specific changes or additions are proposed to achieve this goal. Additionally, it's worth discussing whether there might be any risks or limitations to such an approach that should be considered.

AI-generated content by issue-reviewer may be incorrect