Correctness of classification may leave something to be desired.

Groupings are more or less complete, but there are still ambiguities over transcript biotypes.

Made the circular slice code compartmentalised to just SliceAdaptor (making cache initalisation code easier) and we now only build circular slices when the seqregion it is based off is flagged as circular

Attempting to fix the issues with too many KNOWN status genes appearing in species we should not be seeing them in. UniProtParser now limits to using ACs with an evidence of 1 and 2. We now do not assign KNOWN based upon a depdendent xref when we didn't trust the parent xref. Also the loader & SQL has been edited to allow for LOWEVIDENCE types of sources.

Removing the dead dumper and adding a readme to help with remembering which one is the file we need to work with

Conflict resolved we think. 1.18 code had very odd hash assignment/caching code which should not have worked

Letting users clear the linked caches if they wish to. This is better than users assuming the caches they think they can clear

attempting to fix