Hub shutdown during initialization no longer reported as failure

Short-lived probe hubs (the ones the platform spins up to read a node type's data model and dispose again) could race their own teardown: when a probe was disposed while still initializing, its sub-hubs reported "initialization failed — Hub is now in FAILED state" as errors, and the initialization watchdog fired a second error up to two minutes after the hub was already gone. Both were routine teardown misreported as failures — noise in the error logs with nothing to act on.

Initialization that ends because the hub (or its parent) is shutting down is now recognized as a normal shutdown: no error logs, no failed-state residue, and the watchdog is disarmed when the hub is disposed. A hub that genuinely cannot initialize still fails fast and legibly, exactly as before.