Perform mountr traversals internally in libnfs

[sahlberg]

NFS protocol does not allow a command to traverse across a filesystem boundary.
A consequence of this is that if a server exports two filesystems
/data
and
/data/foo
then the client need to mount them both too since a lookup on foo in the /data directory would result in an object in a different filesystem and thus would not be permitted.

Instead of having the client having to track all sub exports and explicitely jump between mounts when traversing a filesystem tree, we should and can do these special mount handling internally inside libnfs.

This would make using libnfs much more conveneint for consumers when nested mounts are present.

[sahlberg]

done