HFS catalog records are variable length. The checks introduced in
f9e8cbc0 ("Fix for heap OOB memory read issue 1076") and 218a868e ("Check
that the allocated node is large enough to be treated as a
hfs_btree_key_cat.") incorrectly assume that all catalog records have
the maximum possible length, and thus halt processing of a node when
there is still sufficient space for catalog records which are shorter.
This results in many catalog records being missed during filesystem
walks.