If I want to use QNX 8.0 hypervisor in my embedded application, should I put my application code running directly on the hypervisor (native), or should I provision a guest QNX VM and put my code running from the guest QNX VM? Which one is the recommended practice?