Mention new fetch_all_by_GOTerm_accession() method.

  fetch_all_by_child_term() --> fetch_all_by_descendant_term()
  fetch_by_child_term()     --> fetch_all_by_child_term()

  fetch_all_by_parent_term()    --> fetch_all_by_ancestor_term()
  fetch_by_parent_term()        --> fetch_all_by_parent_term()

  fetch_all_by_child_term() --> fetch_all_by_descendant_term()
  fetch_by_child_term()     --> fetch_all_by_child_term()

  fetch_all_by_parent_term()    --> fetch_all_by_ancestor_term()
  fetch_by_parent_term()        --> fetch_all_by_parent_term()


Also, make accessors read-only.

doc corrected

translation_stable_id, translation_version, peptide_archive_id ought
to be NULL for genes that lack a translation.  This patch fixes these
fields to allow for NULLs.

in the gene_archive table to be NULL.

Properly assign internal IDs to terms when loading 2+ ontologies.

the relation_type_id.  SO has one instance where two terms have two
relations between each other.

'part_of' relationships for all ontologies.

Involved tweaks to relation_type loading to invert some relation types
(e.g. 'has_part' => 'part_of') and to allow for sharing relation types
between ontologies.

SO also only has a "default-namespace" defined in the header, but no
"namespace" in the definitions of the terms, so use this.

Fixed a bug in the code that figures out if a file has been processed
before or not.

Use the translations of the _canonical_ transcripts of the source and target genes as the source of GO terms to project, rather than simply the longest translation in each, which was a bit arbitrary and didn't alway work.

Also make the print option for GO term projection a bit more verbose - aids debugging.