[Feat]是否可以支持千问的上下文缓存功能

## Is your feature request related to a problem? Please describe.

千问部分模型可以使用**上下文缓存**功能，但当前项目尚未支持该功能。

## Describe the solution you'd like

我希望能够支持千问的**上下文缓存**功能。具体需求如下：
1. **支持显式缓存**：允许用户在调用千问模型时启用显式缓存。
2. **对外暴露cache_creation_input_tokens**：在使用显式缓存功能时，千问的 API 会返回一个名为 `cache_creation_input_tokens` 的特定字段。我希望系统能够解析这个数值，并在Message.ResponseMeta.Usage中对外暴露。

## Describe alternatives you've considered

暂无

## Additional context

- **千问上下文缓存官方说明文档**: [阿里云百炼 - 上下文缓存 (Context Cache)](https://help.aliyun.com/zh/model-studio/context-cache?spm=a2c4g.11186623.help-menu-2400256.d_0_3_0_8.2fef220fdGKln0&scm=20140722.H_2862577._.OR_help-T_cn~zh-V_1#1ba8a19b9dmui)

The issue names no files or tests, so first locate the existing Qwen integration and its handling of request options and Message.ResponseMeta.Usage. Read the linked Qwen context-cache documentation, then verify that explicit caching and cache_creation_input_tokens are both represented; done means the requested option and usage value are supported.

Written by the indexing model from the issue text.