amd64 pmap: Optimize PKU lookups when creating superpage mappings

[alc]

Summary

Modify pmap_pkru_same() to update the prototype PTE at the same time as
checking the address range. This eliminates the need for calling
pmap_pkru_get() in addition to pmap_pkru_same(). pmap_pkru_same() was
already doing most of the work of pmap_pkru_get().

A similar change was already made to BTI lookups on arm64.

[kib]

sys/amd64/amd64/pmap.c

11493

But don't we need to clear X86_PG_PKRU_MASK first?

[alc]

sys/amd64/amd64/pmap.c

11493

We don't set the PTE field using pmap_pkru_get() in pmap_enter() until after we've already called pmap_enter_largepage() or pmap_enter_pde(). Similarly, pmap_enter_object(), pmap_enter_2mpage(), and pmap_enter_quick() never call pmap_pkru_get(). They rely on pmap_enter_quick_locked() or pmap_enter_pde() to set the PTE field.

As a safety belt, I added the KASSERT("PKU field not set") to the start of pmap_pkru_same().