When deploying Windows Server 2019 VMs in a scale set via an Azure Resource Manager template, which method should be used to ensure NGINX availability on all VMs post-deployment?