add kasserts to nfsm_trimtrailing as suggested by kib@

[rmacklem]

Summary

Add KASSERTS to nfsm_trimtrailing() to confirm the sanity of
the arguments for the M_EXTPG case.

Kostik, is this what you were suggesting?

Test Plan

Set ND_EXTPG flag for all NFS RPCs so the the case
where VOP_CREATE() returns ERELOOKUP hsppens
executing nfsm_trimtrailing() with M_EXTPG mbufs.

[kib]

Yes, intent was to re-check that bpos points to the same position as bextpg with bextpgsiz. But I see (PAGE_SIZE - bextpgsiz) and now see that my understanding was wrong.

Please note that PHYS_TO_DMAP() is not available on all platforms, other code does #ifdef PHYS_TO_DMAP around code that only makes sense on machines with DMAP.

[rmacklem]

Yes, bextpgsiz is the bytes remaining to be used in the page.
It was easier than using offset within the page, since there
is 1st_page_off to deal with.

Actually PHYS_TO_DMAP() is defined for all arches,
but just does a panic() on the ones (like i386)
that don't support it.
--> NFS over TLS is only enabled for arches that

support DMAP and that is the only time the
NFS code generates M_EXTPG pages.

--> Should be ok.