v0.4.2

Fixed

• Fixed a DB checkpoint race where a concurrent pending delete could
  remove an in-flight cache image after the checkpoint worker had
  claimed it, causing write_through_batch: flushing entry lost cache image and blocking crash-safe checkpoint completion.
• Kept pending-delete cleanup from reclaiming cache and route-resident
  state until the delete has been applied to the inner blob store.