If the client loses the connection to the donor, the client can re-run the donorStartMigration command. Currently the retry will return without doing anything.

We should update the donor's logic so that it creates and registers a TenantMigrationAccessBlocker as its first step, and retries call TenantMigrationAccessBlocker::onCompletion to wait for the migration to finish.

This is a preliminary step to moving the donor's logic onto Primary-Only Service.

This ticket should test that a donorStartMigration retry blocks until the migration commits or aborts.