test(e2e): recover temporary error recorders after interrupted runs #380

The journey harness installs a temporary error-recorder plugin before PodNotes so uncaught startup errors fail E2E tests. Normal suite teardown removes its folder and restores the exact enabled-plugin configuration.

An interrupted process can skip teardown. Static inspection of tests/e2e/journey/library.ts shows that the next harness install neither removes stale podnotes-e2e-errors-* folders nor removes stale configuration entries. This risk was identified in independent QA1 review. A SIGKILL interruption was not reproduced during acceptance.

The isolated Linux runner resets enabled plugins during provisioning, and deleting the isolated test vault removes stale folders. The macOS launcher can select the shared dev vault, so recovery should also handle stale entries without changing the user's other plugins.

Add startup reconciliation under the vault lock. Preserve unrelated plugin folders and configuration order. Prove normal rollback and recovery from an interrupted isolated run. Keep the startup recorder ahead of PodNotes on restart.