fix: 修复 kimi-for-coding 开启思考模式无法回复的问题

[Blueteemo]

问题描述

当使用 kimi-for-coding 模型开启思考模式（adaptive/思考预算0/思考深度Low）时，用户无法收到回复，控制台显示有思考内容但无报错。

根因分析

在 tool_loop_agent_runner.py 中，最终结果的 yield 逻辑只检查了 result_chain 和 completion_text，没有处理仅有 reasoning_content 的情况。当模型仅返回思考内容而 completion_text 为空时，最终响应不会发送给用户。

修复方案

在以下两处添加对 reasoning_content 的分支处理：

1. 主逻辑 (721-739行)
2. skills_like fallback (754-767行)

当 completion_text 为空但 reasoning_content 有值时，将 reasoning_content 作为 fallback 返回给用户。

Related Issue

Fixes #7656

Summary by Sourcery

Bug Fixes:

• Return reasoning content as a fallback response when completion text is empty so users still receive replies in thinking mode.

[sourcery-ai]

Hey - I've left some high level feedback:

• Consider extracting the shared llm_result yielding logic for completion_text / reasoning_content into a small helper to avoid duplicating the same AgentResponse construction in multiple branches.
• Double-check that exposing reasoning_content directly to end users is acceptable for all calling contexts, as it may contain internal chain-of-thought or debugging details that were previously intentionally not surfaced.

Prompt for AI Agents

Please address the comments from this code review:

## Overall Comments
- Consider extracting the shared `llm_result` yielding logic for `completion_text` / `reasoning_content` into a small helper to avoid duplicating the same `AgentResponse` construction in multiple branches.
- Double-check that exposing `reasoning_content` directly to end users is acceptable for all calling contexts, as it may contain internal chain-of-thought or debugging details that were previously intentionally not surfaced.

---

Sourcery is free for open source - if you like our reviews please consider sharing them ✨

• X
• Mastodon
• LinkedIn
• Facebook

_{Help me be more useful! Please click 👍 or 👎 on each comment and I'll use the feedback to improve your reviews.}

[gemini-code-assist]

Code Review

This pull request adds support for handling reasoning content in the tool loop agent runner by yielding an AgentResponse when reasoning_content is present. The reviewer suggests explicitly setting the MessageChain type to "reasoning" in both new logic paths to maintain consistency with streaming output and facilitate better UI representation of the thinking process.