[ENSCORESW-577]. Fixing a logical hole where we reqiured a method to be defined but did nothing with it.

Looks like I was between implementations. Now switched to requesting the
additional lookup keys and if the hash was empty then skipping
any additional work. Seems to behave a lot better and means users need
only override one method in the object.

(ie, include the duplicated sequence Y has in common with X rather than the two sequences unique to Y)

DensityType is not a feature but is the object storing the analysis related
to the DensityFeature

unlike other features, the related analysis is stored in the density type table

added test for 'fetch_all_by_feature_class' with density features

Other schemas may need this which do not go through a process of removing
old/retired seq region ids in favour of their new ids. When querying by a slice
this is automatically delt with. When fetching by an ID you now need to
force the search on.

Fixed a warning that was testing the wrong value and failing to verify a store_alt_alleles call properly

It seems that $DBI::errstr has a very strange life. We can get the error message
recived by incorrect connections or bad drivers (DBD::mysql is missing) when we
just warn the error. However concatenating it into another string makes the
message disapear. Even assignment and then concat will not work. Instead we
have switched to using the $@. This has the error in it.

The optimise tree step can take a very long time and cause web timeouts. We have
added a time limit expressed in floting-point seconds. If an optimisation is
taking longer than the specified time we will abort any more attempts.

The optimise tree step can take a very long time and cause web timeouts. We have
added a time limit expressed in floting-point seconds. If an optimisation is
taking longer than the specified time we will abort any more attempts.

The original code decalred a max of 2K ids because of max allowed packet concerns
and wanted the packet to stay close to 256KB. However the actual figure was off
by a factor of 8 as the max packet is 1MB not 1Mbit. We have increased the limit
to 16K when working with numerics. Strings are still at 2K due to their possible
large size.

Fixes for the alt alleles code. We needed to make sure we used the right variables in the code. Also needed to make sure an array is actually defined during construction

Fixing an issue raised by vega and non-standard DBAdaptor group names. Vega use ensembl to denote the ensembl db. We assumed incorrectly that core could always be used. That would always fail the moment a different group name was used

replaced ok tests with is where possible
replaced core test database with patch database

by farm3.

small queue for exonerate to finish.  Also added spaces after -R and -J
switches.

- get_assembly_name
- get_assembly_date
- get_genebuild_start_date
- get_genebuild_method
- get_genebuild_initial_release_date
- get_genebuild_last_geneset_update

bug fixes for methods:
- get_toplevel
- get_karyotype
- get_coord_systems

returns a listref of all unique assembly versions available

will return a single string with the value of the default version of the assembly

replaced by GenomeContainer, hopefully more generic and in line with what already exists
presenting most of the generic features the assemblyadaptor add, but less REST only focused
will be further expanded with a genome table

This work allows for the hosting of more than one variation database (similar to
core-like databases such as rnaseq or otherfeatures). The Slice methods for core objects
allowed the specification of the database type. Variation methods now have the same
option. We have pushed much common logic into 4 methods to provide this automatic
defaulting behaviour.

We needed more methods to be able to handle alt allele group interactions. We also
needed to make sure a single gene could not be added more than once. Think we got
this one fixed