unused_oids script is broken with bsd sed

Hi.

BSD sed in macOS doesn't understand word boundary operator "\b". So after
372728b0d49 unused_oids doesn’t match all oids in new *.dat files marking
all of them as unused.

It is possible to write more portable regexps, e.g. change "\b" to
something like "^.*{*.*", but it seems easier for feature use to just
rewrite unused_oids in perl to match duplicate_oids. Also add in-place
complain about duplicates instead of running uniq through oids array.

--
Stas Kelvich
Postgres Professional: http://www.postgrespro.com
The Russian Postgres Company