Add ability to load from previously exported results, so IDs can be saved and rereferenced.

[SecretSheppy]

The main problem with this is how we access the source files after the results have already been processed. We could quite easily store all the source files in a json, which would make this much simpler.