Tasks: a readable session stream, a session tree, and a chat bar

Phase 3 of 3. A DAG, not a list: tasks with the same dependencies can run in any order. Each _Test:_ names a row of testing-plan.md.

• [x] 1. transcriptThread projection — replace transcriptTurns in ui/src/api/model.ts with the paired, summarised row model from the design (R1, R2.1). Test: T1.
• [x] 2. components/Transcript.tsx — TranscriptView (collapsed <details> tool rows, thinking, meta) and ChatBar (send / disabled-with-reason) (R2, R4.1, R4.4). Test: T3.
• [x] 3. Sessions view — views/Sessions.tsx, #/sessions[/<ref>] route, Nav tab; sidebar from sessionTree (R3). Test: T2, T8.
• [x] 4. WorkItemDetail reuse — trace panel renders TranscriptView; ChatBar bound to the selected trace tab; delete the superseded TurnRow markup. Test: T3.
• [x] 5. Projection tests — pairing, orphans, thinking, meta, malformed, summaries (T1 rows) in model.test.ts. Test: T1.
• [x] 6. Reply resolves PR endpoints — reply_session via record_owning/endpoint_for with the closed-endpoint fallback; refusals kept (R4.2, R4.3). Test: T4.
• [x] 7. Reply route tests — PR-ref delivery, fallback, kept refusals in cli/tests/test_ask_reply_integration.py. Test: T4.
• [x] 8. OpenAPI prose — /sessions/reply description says a PR ref resolves to its endpoint. Test: T7.
• [x] 9. Demo fixture — transcript with ids + paired results + thinking + meta; an ad-hoc work item; titles (NFR3). Test: T8.
• [x] 10. Component tests — Transcript.test.tsx for collapsed/expanded rows and chat-bar states. Test: T3, T12.
• [x] 11. sessionTree tests — tree shape and ad-hoc flag in model.test.ts. Test: T2.
• [x] 12. Verification + evidence — run the plan, fill its verification record, commit evidence. Test: T5, T6, T14.
• [x] 13. Capability + user docs — docs/capabilities/control-plane.md, ui/README.md, execution log's Documentation section.