Slight overhaul of ISBN indexing focused on the translate_isbn1013 function

Slight overhaul of ISBN indexing focused on the translate_isbn1013 function

  * Switch to a PLPERLU function built on Business::ISBN for more robust
    indexing of records with problematic ISBNs; thanks to Jason Stephenson
    <jstephenson@mvlc.org> for the initial implementation

  * Remove the first_word, naco_normalize, and split_date_range normalizers
    from the indexing chain for ISBNs, as these unnecessarily munge the
    ISBNs. We can trust Business::ISBN to determine what is a real ISBN,
    hyphens and all.

  * Index all ISBNs in a given record and generate the corresponding
    ISBN10/ISBN13 variations. Also, fix the checksum if given an ISBN with
    a bad checksum, but be sure to index the original bad-checksum ISBN as
    well.

  * Add a set of 10 MARC records with bad ISBNs (thanks to Jason Stephenson
    again for these) for testing purposes to Open-ILS/tests/datasets

  * Add the Business::ISBN Perl module as a prerequisite to the installer

  * Remove contrib/isn from the PostgreSQL install instructions

git-svn-id: svn://svn.open-ils.org/ILS/trunk@17066 dcc99617-32d9-48b4-a31d-7c20da2025e4