Re-introduced hack for handling zero-sized entries as explained in e731ff2.
Do not ignore corrupted entries on iteration because this behavior prevents eblob self-healing to kick in and "fix" corrupted index; Also those corrupted entries then go to sorted index that can prevent blob from starting up next time.
Currently "fix" means simple ftruncate(2) of index file to the last non-corrupted entry. See: http://doc.reverbrain.com/roadmap:eblob
ftruncate(2)
of index file to the last non-corrupted entry. See: http://doc.reverbrain.com/roadmap:eblob