Implement per-CPU variables via CPU-local storage (#917)

* Only supported on x86_64 currently; aarch64 support to come.

* x86_64 implementation is based on the GS segment register
  and the GsBase MSR, as typically used on other OSes.

* Tested working, but not yet used anywhere in Theseus.

Co-authored-by: Klim Tsoutsman <klim@tsoutsman.com> fd5081a94923c4408cb40c326dbbbaec4d4887d3