Use the selected chat model for the Ruby design agent

[Hashim1999164]

Summary

• The design chat agent always asked vscode.lm.selectChatModels for Copilot GPT4o, so custom chat models left model undefined and sendRequest crashed
• It now uses the model already selected for the chat, falls back to Copilot GPT4o if needed, and shows a message instead of throwing when no model is available

Closes #4186

Tests

• npx tsc covering the vscode extension

[vinistock]

It looks like both linting and formatting are failing. If request.model is new, it may require bumping the vscode types package to the version that introduces it along with the VS Code engine declared in the package.json.

[Hashim1999164]

Bumped the VS Code engine and @types/vscode to 1.96 so request.model is typed, and put the missing model message on one line for Prettier. lint_node and build_node should be green on this push.