fix(core): fail-fast on corruption, preserve buffer on transient errors

- Throw from init() when integrity check fails and recovery doesn't help,
  preventing writes to a known-corrupt database
- Only drain buffer on successful flush; requeue valid entries on transient
  failures (busy/IO) so they aren't silently lost
- Add spy on flushAgentLogBuffer in deleteTask test to prove flush-before-delete

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

packages/core/src/__tests__/store.test.ts

@@ -4917,8 +4917,11 @@ Task with acceptance criteria
         const task = await createTestTask();
 
         await store.appendAgentLog(task.id, "to be cascaded", "text");
-        // deleteTask should flush first, then cascade-delete the entry
+        // Prove flush happens before delete
+        const flushSpy = vi.spyOn(store as any, "flushAgentLogBuffer");
         await store.deleteTask(task.id);
+        expect(flushSpy).toHaveBeenCalled();
+        flushSpy.mockRestore();
 
         const after = (store as any).db.prepare(
           "SELECT COUNT(*) as count FROM agentLogEntries WHERE taskId = ?",