amd64: switch pmap_map_io_transient() to use pmap_kenter_attr()

instead of constructing transient pte itself.  This pre-set PG_A and
PG_M bits, avoiding atomic pte update on access and modification.  Also
it set the nx bit, the mapping is not supposed to be used for executing.

amd64 pmap: assert and explain why pmap_qremove() is safe WRT supermappings

Based on alc@ comments from https://reviews.freebsd.org/D47678.