Once the nodemap/idmap changes are landed, it makes sense to replace the existing root squash code with equivalent handling via the default_nodemap instead of having two overlapping mechanisms.

That would simplify the Lustre code a bit, and get better testing of the nodemap code, which has a superset of functionality. It makes sense to keep the existing /proc interfaces for setting the nosquash NIDs and UID/GID, and just have them modify the default_nodemap structure appropriately.