After talking to Compara we found the need to support additional
lookups with the full id cache. Since all values are stored in memory
the lookup is very easy to compute.

The production and web teams have reported odd going ons when people copy from
staging onto another server. This is because this script optimises databases
before copying. These DBs should already be optimised so there is no need
to do this stage. In fact it can alter the checksums of the MYI files which
causes the webteam to wonder if data has been switched out without them knowing.

Leo submitted patch. We support pgsql as a way of defining the driver. This
is inline with us accepting sqlite as a valid synonym for DBD::SQLite.

To provide improved querying of caches we now support SQL statements. These
queries return a list of keys (normally primary key db identifiers) defining
a list of objects to return. It is up to sub-BaseCache implementations to
deal with the nitty gritty of cache mis-hits (something they need to do
anyway).

- seq: pass defined slice intervals to trigger the correct program flow in SequenceAdaptor
- project: removed since believed to have bugs
- project_org: removed since it's not used anywhere

The code previously allowed bogus numbers through into the final slice which
caused errors in later methods. The website crashed because of this. We will
reject anything that was not an integer if the start or end was defined.

Nathan requested methods which can perform this kind of task (assert the contents
of a data structure). We support 1 level of arrays and hashes. All test cases
pass at time of writing. Methods for capturing method returns were
not implemented. They do not fit into the Utils way of doing (no method
actually performs a task on an object they just report what it is).

also some ergonomical changes:
- get a count+ is test replaced by is_count method
- variable names declared at the top of the modules, so can easily change gene/dbid/db used if the test database changes

Changed expected GTF in some cases: the start/stop codons should not have been there (incomplete CDS annotation)

Removed get_all_ExternalFeatures method, now borrowed from parent where the corresponding adaptor handle the circular slice case

Removed some Slice overridden get_all_(Feature) methods since features/slices spanning the origin of replication are handled the corresponding feature adaptor

Modified the enum on object_xref from 72->73 patch c and the redump
re-did some data. Such as choosing a different encoding for some very
small values. All changes are small though.

now fully testing patches and haplotypes