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