Efficient fetching and handling of annotation across genomes and proteomes

The Jalview 2.x series datamodel keeps independent dataset sequences for DNA and protein products, and annotation is independently maintained on both sequence objects.

Large scale feature retrieval operations take considerable time - so it must be made as efficient as possible:

* datasources should be queried once for any identifier - repeated retrievals should be avoided unless the datasource has updated information about the identifier.

* Where there are unambiguous relationships between sequences (e.g. known protein products of coding regions), then any retrieved or manually created annotation should be shared between all forms of the sequence data.