Add resolved model for events

[dennyac]

Add resolved model to user action telemetry

• When modelId is copilot/auto, replace it with the actual model that served the request
• Leverages existing resolvedModel field from the server response, already plumbed in Add resolved model #1517
• Surfaces resolvedModel in IResultMetadata so downstream consumers can access it
• Affects events: conversation.acceptedCopy, conversation.acceptedInsert, conversation.appliedCodeblock
• Non-auto model selections are unchanged

[Copilot]

Pull request overview

Adds support for recording the resolved model (the actual model that served a request) in user-action telemetry, especially when the user selected copilot/auto, by propagating the server-provided resolvedModel through result metadata and using it in the affected telemetry events.

Changes:

• Plumbs resolvedModel into IResultMetadata and populates it from successful ChatResponse results.
• Updates user action telemetry to substitute copilot/auto with the resolved model for copy/insert/apply codeblock events.
• Updates the static chat fetcher test helper to allow configuring resolvedModel on responses; adds a new test file.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
src/platform/chat/test/common/staticChatMLFetcher.ts	Allows test fetcher responses to return a configurable resolvedModel.
src/extension/prompt/node/defaultIntentRequestHandler.ts	Copies response.resolvedModel into the metadata fragment on successful responses.
src/extension/prompt/common/conversation.ts	Extends IResultMetadata with optional resolvedModel.
src/extension/conversation/vscode-node/userActions.ts	Replaces telemetry modelId with metadata.resolvedModel when the action model is copilot/auto.
src/extension/conversation/vscode-node/test/userActionsResolvedModel.spec.ts	Adds tests, but currently re-implements the production mapping logic instead of exercising it.

Comments suppressed due to low confidence (1)

src/extension/conversation/vscode-node/userActions.ts:281

• The copilot/auto resolution logic and fallback string are duplicated inline here and again in handleApplyAction. To reduce the chance of future drift, consider extracting a small helper (and a shared AUTO_MODEL_ID constant) that maps (requestedModelId, resolvedModel) to the telemetry modelId.

					languageId: e.action.languageId ?? '',
					modelId: e.action.modelId === 'copilot/auto' ? (result.metadata?.resolvedModel || 'copilot/auto') : (e.action.modelId ?? ''),
					comp_type: compType,

[Copilot]

There’s no unit test asserting that resolvedModel is actually propagated into chatResult.metadata from a successful ChatResponse. Since defaultIntentRequestHandler.spec.ts already covers getResult(), it would be better to add an assertion there (e.g., set the test IChatMLFetcher to return a non-empty resolvedModel and verify it appears in the returned metadata).

[dennyac]

Fixed

[dennyac]

Tagging @lramos15 as it's related to #1517