fix(#2995): ignore leading vis-thinking block in react parser

[LXW2019124]

Description

更新了react_parser.py，让 ReAct 解析在提取 Thought / Phase / Action / Action Input 之前，先忽略前置的 vis-thinking 包装内容。

这次改动只发生在 parser 层，范围很小：
当模型输出以前置 VisThinking 块开头时，parser 会先剥离这段 thinking 包装，再继续解析正文。
如果剥离之后没有正文内容，parser 会返回空步骤列表，而不是把 thinking 内容误当成可执行的 ReAct step。
thinking 结束位置的判断也做了收敛，避免两类边界问题：
不会因为 thinking 内部出现 fence 而过早截断
不会把后续真实正文错误吞进 thinking 区域
这个修改修复了 deepseek-reasoner 在 ReAct 场景下，因 reasoning 内容被误解析成多个 step，从而触发 Only one action is allowed each time. 的问题。

How Has This Been Tested?

测试覆盖的场景包括：
标准 ReAct 文本解析行为保持不变
前置 vis-thinking 块会被正确忽略
只有 vis-thinking、没有正文时，返回空步骤
thinking 内容内部包含 standalone fence 时，仍能正确跳过 thinking
正文本身包含 standalone fence 时，不会被错误吞掉，仍能正常解析

Snapshots:

Include snapshots for easier review.

Checklist:

• My code follows the style guidelines of this project
• I have already rebased the commits and make the commit message conform to the project standard.
• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• Any dependent changes have been merged and published in downstream modules

[Aries-ckt]

I am very appreciate your contribution. we will test and merge it soon.

[chenliang15405]

Thanks for your contribution. LGTM

[Aries-ckt]

r+