Add OTel metrics for edit acceptance and code survival telemetry

[digitarald]

Summary

Adds OpenTelemetry counters and histograms for edit acceptance and code survival across inline chat and chat editing sessions, supplementing existing MSFT/GH telemetry.

Changes

New OTel metrics (4 instruments):

• copilot_chat.edit.acceptance.count — counter for accept/reject of inline chat and chat editing hunks
• copilot_chat.chat_edit.outcome.count — counter for chat editing session outcomes (accepted/rejected/saved)
• copilot_chat.edit.survival.four_gram — histogram for 4-gram text similarity survival scores
• copilot_chat.edit.survival.no_revert — histogram for no-revert survival scores

GitHub telemetry additions:

• sendGHTelemetryEvent for inline.done and inline.trackEditSurvival (previously MSFT-only)

Type safety:

• EditSource and EditOutcome union types enforce valid attribute values at compile time
• All attribute keys use copilot_chat.* namespace via CopilotChatAttr constants

Files changed

• genAiAttributes.ts — attribute constants + union types
• genAiMetrics.ts — 4 new static metric methods
• userActions.ts — wiring at 4 call sites + GH telemetry
• agentTraceHierarchy.spec.ts — 2 new tests (happy path + edge case)