Change dataset identifiers from random numbers to dataset urls

[alexsb]

Currently the datasets are assigned a number based on when they are loaded. The parallel loading, however, makes this numbering undeterministic, thus we can't link directly to a dataset.

We should replace this number with the url of the json file.