chore: add changeset for v3.26.5

[mrubens]

Important

Add features and a fix in v3.26.5, including new model support, auto-approve, and performance improvements.

• Features:
  • Add support for Qwen3 235B A22B Thinking 2507 model in chutes.
  • Add auto-approve support for MCP access_resource tool.
  • Add configurable embedding batch size for code indexing.
  • Move message queue to the extension host for better performance.
• Fixes:
  • Add cache reporting support for OpenAI-Native provider.

^{This description was created by for 642abe7. You can customize this summary. It will automatically update as commits are pushed.}

[roomote]

Thank you for preparing the v3.26.5 release changeset! The changelog entries are clear and properly reference the related issues. I've left a few suggestions for minor improvements that could help maintain consistency in future releases.

Additionally, I noticed the release image (releases/3.26.5-release.png) is 1.6MB - as releases accumulate over time, these large images might significantly increase the repository size. Consider using image optimization tools or a more compressed format while maintaining visual quality.

[roomote]

Is using the version number as the changeset filename intentional? Typically, changesets use randomly generated names (e.g., 'brave-pandas-dance.md') to avoid conflicts when multiple changesets are created for the same version. This approach might work for release changesets, but wanted to check if this follows your established pattern.

[roomote]

Could we standardize the attribution format for better consistency? I notice some entries use 'thanks @username!' while others use 'by @username, PR by @username'. A consistent format would make the changelog more uniform.