feat(telemetry): add autoresearch usage tracking

[bbsngg]

Summary

• Adds dedicated autoresearch_* telemetry events covering the full usage funnel: hub discovery → pack/workflow selection → configuration → message submission → backend run lifecycle
• New helper src/utils/autoresearchTelemetry.ts with micro-batched queue (reuses existing /api/telemetry/events pipeline and isTelemetryEnabled() opt-out)
• Backend server/routes/auto-research.js emits autoresearch_run_started, _task_completed, _finished (with outcome + duration), and _cancel_requested via enqueueTelemetryEvent
• Backfills data-telemetry-id on autoresearch buttons so existing ui_click events gain semantic identifiers as a backup data lane

Why

Existing telemetry only captures generic DOM events (ui_click, ui_change, ui_submit) plus agent_dialogue_meta. It cannot answer core product questions: which pack are users picking? Which workflow is most used? Does the config flow convert? Are autoresearch runs finishing or failing? This PR closes that gap with a focused event set.

Privacy

• User input content is never sent — only input_length, command (from scenario_id), and attachment_count
• project_path is SHA256-hashed (first 16 chars) before emission
• API keys / env var values are dropped; only key names and filled-count are reported
• Server-side sanitizer (server/routes/telemetry.js BLOCKED_EVENT_KEYS) continues to strip content / prompt / output / message / context as a second line of defense

Events added

Frontend (source: autoresearch-ui): autoresearch_hub_opened, _hub_viewed, _dropdown_toggled, _pack_expanded, _workflow_selected, _scenario_clicked, _config_section_opened, _mcp_selected, _configure_clicked, _configure_result, _guide_toggled, _guide_dismissed, _message_sent

Backend (source: auto-research-backend): autoresearch_run_started, _run_task_completed, _run_finished (outcome: completed/failed/cancelled + duration_ms), _run_cancel_requested

Test plan

• npm run typecheck passes
• npm run build passes
• Backend module imports without errors
• Manual: start dev server, open Auto Research hub via sidebar, expand a pack, select a workflow in chat dropdown, verify POSTs to /api/telemetry/events contain the new event names and no user content
• Manual: toggle Settings → Data Transfer off, repeat above, verify no requests are sent
• Manual: trigger a full autoresearch run (start → task completes → finish), verify backend emits autoresearch_run_started and _run_finished with correct outcome
• Manual: cancel an active run, verify autoresearch_run_cancel_requested is emitted

🤖 Generated with Claude Code