Add support for vGPU for instance with accelerator

Currently, in Nova, we can enable GPU devices (`enabled_vgpu_types`) in nova.conf , and add "resources:VGPU : 1" [1] metadata to the flavor, then we can boot instance with this flaovr. For this way, we just can allocate one vGPU for the instance, and just support boot instance one by one, it's not friendly for users.

We would like to support vGPU by Cyborg, and we can boot instance with vGPU. Make the pGPU and vGPU managed by Cyborg.

[1]https://docs.openstack.org/nova/ussuri/admin/virtual-gpu.html#configure-a-flavor-controller