cdna_coding_end(), coding_region_start(), and coding_region_end() on
exons.

Current API and test DB passes the tests.

    $exon->cdna_start($transcript);
        # Returns the start position of the exon in cDNA coordinates.

    $exon->cdna_end($transcript);
        # Returns the end position of the exon in cDNA coordinates.


    $exon->cdna_coding_start($transcript);
        # Returns the start position of the coding region of the exon
        # in cDNA coordinates.  Returns undef if the whole exon is
        # non-coding.

    $exon->cdna_coding_end($transcript);
        # Returns the end position of the coding region of the exon
        # in cDNA coordinates.  Returns undef if the whole exon is
        # non-coding.


    $exon->coding_region_start($transcript);
        # Returns the start position of the coding region of the exon in
        # genomic coordinates on the forward strand.  Returns undef if
        # the whole exon is non-coding.

    $exon->coding_region_end($transcript);
        # Returns the end position of the coding region of the exon in
        # genomic coordinates on the forward strand.  Returns undef if
        # the whole exon is non-coding.


Since an exon may be part of one or more transcripts, the relevant
transcript must be given as argument ot these methods.

Changed NULL in output files to \N so that it's recognised as a "proper" null value when imported into MySQL.

score is no longer an attribute of ArchiveStableId (this is now handled by StableIdEvent); history tree consolitation and untangling calls

Corrected documentation for created_date() and modified_date() - they actually use a UNIX time integer value, not a MySQL date.

Fixed typo (= instead of eq) that was causing search regions to be deleted when enhancers were loaded!

Prevent race condition which could lead to confusing results if two projections to the same target (e.g. human-dog and mouse-dog) were run in parallel.

commented out by someone foolish).

changes from code review implemented. Principally the reduction of some duplicated code. Also for the generation of display_xref_ids do not use those where the display_label is just a number. Also perl -w now runs with no warnings. Ah yes and also the triage stuff is now done right at the end as this was done before the pair calculation before which made writing the xrefs for the pairs more complicated than was needed.