Merging of UniParc mapping code into the Xref system. This relies on being able to contact the EBI based UniParc readonly database.

Brought in new_fast() as a speed increase.  Should probably consider moving new_fast() up to Storable.pm to give it to as many modules as possible & reduce code duplication

; TODO ; do_similarity_scoring      = 0
; TODO ; do_project_to_common_cs    = 0

Faster version of rearrange which ditches uc() and order modifying in favor of editing the input parameters directly with a tr///d. This is 35% faster than our original implementation. Test case suite has been run before this commit.

Caches are now created from an accessor rather than building them in the new(). This allows us to remove the new() and lets people inherit from here however they want to. It also stops us from auto-vivifying a plain hash which causes unintentional object retention.

clear_cache() did not de-reference the hash before clearning which could have caused the same situation of creating an auto-vivified hash.

Regression test put in place to stop this from happening again.

(--species) and -t (--type).  Not yet implemented.

Adding a global here to allow for silencing of the no_cache warning. Important for dump applications where caching is not that useful

This is an option for turning on or off exonerate.  You might want to do
this if the assembly hasn't changed.

Use seq_region mappings correctly by getting all external seq_region_ids
for our internal seq_region_id and then using an IN-list in the
subsequent query.

add_xref now takes a hash instead  of a list for easy maintainance. Lots of docs and removal of old subroutines. perlcritic fixes.