Preprocess coordinates by merging adjacent inserts and deletes (mis-matches) so that they are treated like matches. Lose fewer exons this way.

Added methods: hseq_region_name, hseq_region_strand, hseq_region_start and hseq_region_end, that return something if hslice is defined

* store xrefs to human genes/transcripts/translations that chimp genes were generated from

moved routines related to storing of final genes into Gene package, added splitting of v. long introns, clustering of transcripts into genes, removal of duplicate transcripts and brinnging   across xrefs

addition of a $sth->finish

chained mapper was already being used even though faster regular mapper could usually have been used (accidentally commited with cache fix commit)

split input of seq_region files into seperate query/target seq_region files because the same names may be used on both sides

removed phase recalculation code, instead retain human phases which is the correct thing to do when exons are dropped

Modified store method slightly.  display_xref is now stored if set so update() does not need to be called as seperate step

Made global seq_region cache non-global.  Global cache for all dbs + mod_perl = very mysterious web bug.

In deprecated chr_name call, I've changed the code so it projects to the toplevel, not to the chromosome.  All species used to have chromosomes, so this always worked - the meaning of this call is now the same as it was, with the addded bonus that it works for species without chromosomes

Moved meta_coord table information retrieval/storage out of CoordSystemAdaptor and into new MetaCoordContainer class.  This was causing problems because the meta coord information should be stored/retrieved in the feature database and the coord system information should be stored/retrieved in the dna database