v2: fix agent-runner lifecycle and session DB reliability

- Use DELETE journal mode for session DBs instead of WAL. WAL doesn't
  sync reliably across Docker volume mounts (VirtioFS), causing dropped
  writes and duplicate deliveries.
- Add 20s idle detection to end the query stream. The concurrent poll
  tracks SDK activity via a new 'activity' provider event. When no SDK
  events arrive for 20s and no messages are pending, the stream ends
  and the poll loop continues.
- Add touchProcessing heartbeat so the host can distinguish active
  agents from idle ones by checking status_changed recency.
- Catch query errors in the poll loop and write error responses to
  messages_out instead of crashing the process.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

container/agent-runner/src/poll-loop.test.ts

@@ -120,10 +120,11 @@ describe('mock provider', () => {
       events.push(event);
     }
 
-    expect(events.length).toBeGreaterThanOrEqual(2);
-    expect(events[0].type).toBe('init');
-    expect(events[1].type).toBe('result');
-    expect((events[1] as { text: string }).text).toBe('Echo: Hello');
+    const typed = events.filter((e) => e.type !== 'activity');
+    expect(typed.length).toBeGreaterThanOrEqual(2);
+    expect(typed[0].type).toBe('init');
+    expect(typed[1].type).toBe('result');
+    expect((typed[1] as { text: string }).text).toBe('Echo: Hello');
   });
 
   it('should handle push() during active query', async () => {