Since commit 3d36e57f ("gfs2: gfs2_create_inode rework"),
gfs2_evict_inode() and gfs2_create_inode() / gfs2_inode_lookup() will
synchronize via the inode hash table and we can be certain that once a
new inode is inserted into the inode hash table(), gfs2_evict_inode()
has completely destroyed any previous versions.  We no longer need to
worry about overlapping inode object lifespans.  Update the code and
comments accordingly.
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>