should validate hardcoded kernel structure info

in #7837 I added this mild crime of just writing in the size of a page_t in service of estimating how much memory OS allocations would account for in a fully-loaded sled. if page_t changes, this will become wrong.

from a quick skim, there's also dk_minfo_ext which is quite stable, but we don't have anything to validate this (or other, if there are other kernel structs we use in Omicron) is actually correct.

in the page_t case it hasn't changed in 20 years, and I see no reason to expect it to change soon. this is very low priority. but we tossed some ideas around and tried some stuff in this conversation thread - ctest2 has some wrinkles here, walking CTF at runtime might be better, or if we extract illumos bindings into a more standalone crate that might be more amenable to ctest2 as well.

either way this is pretty low priority.

Start with the hardcoded page_t size in sled-hardware/src/lib.rs and the dk_minfo_ext definition in illumos-utils/src/dkio.rs. Review the linked discussion on PR #7837, including the ctest2 and runtime CTF ideas, before choosing an approach. Done means kernel structure information used by Omicron is validated against the corresponding illumos definitions.