Description of problem

Currently the VM rootfs can be a nvdimm or a initrd. We should allow pure block based rootfs.
In the case of firecracker we currently use a block based VM rootfs passed in via virtio-mmio.
However we could used block based rootfs (virtio-blk, virtio-scsi or virtio-mmio) even with QEMU.

This will impact the block device addition logic which does not account for the the fact that the VM itself would consume a block device.

The code should also be factorized to treat the VM block rootfs to be a regular drive. This will also lead to the kernel command line being appended with root=/dev/xyz where xyz is generated based on type of rootfs.