Adding in methods to retrieve repeat types from repeat_consensus. Also increased the number of types we can limit by at once

Changes to allow for a global path setting to be given as well as delegating to the meta table. Also made the data_file be able to hand back the external data adaptor

Bugfix for issue raised by Michael Nuhn from Ensembl Genomes. fetch_all_by_Transcript_list() incorrectly assigned the canonical translation ID

Introduction of new FASTASerializer to replace SeqDumper's FASTA capability. Other formats remain SeqDumper's responsibility.

is_canonical() can lazy-load the value from the database when it is unset & a dbID and an adaptor is available.

has unrelated remark beginning with string "selenocysteine ". Found to be the
case in OTTDART00000030742 for Dec 19th 2011 Havana data-freeze of Zebra fish.

First stage commit only. Partial functionality (detailed below), but works and reduces
chance of missing problems.

1. Made exception data structure richer for check_for_stops to allow a number of different
types of exception to to be expressed.

2. Added a new exception type, that of known TGA-stopping Sec-coding transcript.

3. Added data-structure to add_selcys stop checking to allow specification of known
TGA-stopping Sec-coding transcripts.

1. Fix check_for_stops to cope with multiple adjacent internal stops without a loss of codon to AA sync.
Found when running add_selcys.pl on zebrafish for Vega 47.

2. Fix to existing conditional in vega Sec QC script which tests for an anacode bug before deciding to complain
about missing Sec markup for selenoproteins that also stop with TGA, so that this conditional actually spots
this case in more proteins and so cries wolf less often.

unique genes method (for assembly patches), interactive does have an argument, other binaries default to 1 if nothing specified

toplevel location code can now deal with excessively large regions and shrinks them down to size if possible or rejects them.