Update OpenAI Responses API compaction and telemetry

[dileepyavan]

Summary

This PR refines how the OpenAI Responses API context-management compaction feature is enabled and observed, ensuring compaction-specific behavior only runs when compaction is actually enabled and adding outcome telemetry plus regression coverage.

Details

• Ensure the previous responseId is passed when ignoreStatefulMarker is false.
• Centralize computation of the compaction threshold and use it to gate request context_management and WebSocket stateful-marker slicing.
• Propagate the compaction threshold into Responses API stream processors and emit responsesApi.compactionOutcome telemetry only when compaction is enabled.
• Add regression coverage for enabled and disabled compaction flows and telemetry emission behavior.
• Cherry-pick the implementation from Update to Responses API compaction and telemetry vscode-copilot-chat#4931.

Testing

• Not run in this repo as part of this push/PR creation.

[Copilot]

Pull request overview

This PR refines the OpenAI Responses API “context management / compaction” integration so that compaction-specific behavior (request shaping, stream processing, and telemetry) only occurs when compaction is actually enabled, and adds regression tests around the enabled/disabled flows.

Changes:

• Centralizes compaction threshold computation and uses it to gate context_management request fields and compaction outcome telemetry.
• Improves stateful-marker + compaction round-tripping so the latest compaction item is preserved even when it predates the stateful marker.
• Adds unit tests covering request-body shaping, compaction reconciliation, and telemetry emission behavior.

Show a summary per file

File	Description
extensions/copilot/src/platform/networking/node/chatStream.ts	Adds a dedicated responsesApi.compactionOutcome telemetry helper + GDPR annotation.
extensions/copilot/src/platform/endpoint/node/responsesApi.ts	Centralizes threshold logic, updates request shaping, reconciles latest compaction item in stream processor, emits compaction outcome telemetry when enabled.
extensions/copilot/src/platform/endpoint/node/chatEndpoint.ts	Threads computed compaction threshold into processResponseFromChatEndpoint.
extensions/copilot/src/extension/prompt/node/chatMLFetcher.ts	Passes compaction threshold (derived from request body) into the Responses stream processor for WebSocket flows.
extensions/copilot/src/extension/externalAgents/node/oaiLanguageModelServer.ts	Same as above for pass-through external agent streaming.
extensions/copilot/src/platform/endpoint/node/test/responsesApi.spec.ts	Adds regression coverage for compaction enablement, stateful-marker slicing, compaction round-trip, and telemetry emission.

Copilot's findings

• Files reviewed: 6/6 changed files
• Comments generated: 2