Download datasets in a unique path

[aperrin66]

This is necessary because for some datasets, the file doesn't contain the date, so caching a file prevents from downloading files with the same name in the future. Datasets are now downloaded in a folder named using the entry_id of the dataset, ensuring uniqueness.

[aperrin66]

Various other changes:

• updating the converter version for the same kind of naming problem in the files resulting from the conversion
• adding django settings for tests
• adding missing unit tests

[aperrin66]

FYI @akorosov