Always send reasoning for same responses model

[chrmarti]

No description provided.

[Copilot]

Pull request overview

This PR updates Copilot tool-calling prompt rendering to include “thinking/reasoning” data in historical rounds when using the Responses API and the historical round was produced by the same model, by tracking a per-round model identifier.

Changes:

• Add modelId to IToolCallRound / ToolCallRound and populate it from the tool-calling loop.
• Render thinking data for historical rounds only when apiType === 'responses' and round.modelId matches the active endpoint model.
• Add unit tests covering thinking inclusion/exclusion for historical vs non-historical rendering.

Show a summary per file

File	Description
extensions/copilot/src/extension/tools/node/test/toolCalling.spec.tsx	Adds tests validating when historical thinking is included/dropped.
extensions/copilot/src/extension/prompts/node/panel/toolCalling.tsx	Adjusts ChatToolCalls rendering logic to conditionally include thinking (and gates phase by modelId).
extensions/copilot/src/extension/prompt/common/toolCallRound.ts	Renames/stores per-round model identifier (modelId) on ToolCallRound.
extensions/copilot/src/extension/prompt/common/intents.ts	Updates IToolCallRound schema to replace phaseModelId with modelId.
extensions/copilot/src/extension/intents/node/toolCallingLoop.ts	Populates modelId for each produced ToolCallRound.

Copilot's findings

• Files reviewed: 5/5 changed files
• Comments generated: 3