we already thought about this (latest Kernel and latest KVM) but this split will double a lot of work here and this means development and releases will be much slower and noone wants this now (before 2.x).
(but theoretically it could be possible to compile a new Kernel without openvz and use it on Proxmox VE)