Hardware acceleration for unikernels

Early work on porting the vAccel hardware-acceleration framework to the rumprun unikernel, detailing the semantic abstraction for ML inference, the frontend/backend driver port on rumprun over QEMU/KVM, and a demo of unikernels performing image ML inference.