Currently, recipient shards learn they are participating in a resharding operation by detecting reshardingFields during a metadata refresh. This ticket should replace that pull-based mechanism with a coordinator-driven _shardsvrReshardRecipientInitialize command sent during kPreparingToDonate.

Recipient shards will construct and persist their state document, initialize their in-memory state machine, and ensure idempotency. The coordinator will handle command retries to support resilience in the presence of failures.