Bench on KVM: will-it-scale for CortenMM and the huge page flush #533

These items need a host with /dev/kvm or real hardware. R4 had no such host, so they move here from R4.

• Run fault_mt, mmap_mt and will-it-scale on the VMA backend and on the corten backend, on x86_64 and arm64 (from CortenMM: remove the big kernel lock, keep the status in the page table, fast fork #405). Then decide open question 1 (Open question 1: CortenMM or VMAs as the source of truth for address spaces #22) with these numbers.
• Run the flush of one huge page on KVM (from hugetlb follow-up: rebuild the shared LTP tree, hugemmap24 on arm64 and the cow benchmark #435 item 3).

The maps and smaps diff over the L4 corpus is done (#529), and its results are in tools/bench/cortenmm.md.