Add support for mocking vfio gpu devices to test KubeVirt workloads

Currently the driver mocks GPUs devices. However, with kubevirt, once the device allocation is done for the pod, the pod then configures libvirt domain xml and passes the device to the domain (Kubevirt VM guest).

For this reason, currently with the driver, a kubevirt VM never goes to running state. There is a need to create a mock vfio device, which will help run example workflows as well as e2e tests for KubeVirt VM.

Start by tracing the driver's existing GPU-device mock and the device-allocation path used for KubeVirt workloads. The work is complete when a mock VFIO device allows the example workflow and end-to-end tests to reach a running KubeVirt VM guest.

Written by the indexing model from the issue text.