Move filename generation outside if-else block in datasets.py

[magdalendobson]

In datasets.py, when calling prepare on a slice of a dataset, the filename was incorrectly being set to the base dataset filename when checking whether the file already existed. This caused download of smaller slices to be skipped if the base dataset file already existed in the user's data directory. This PR changes the filename to the cropped filename before checking for existence and thus makes sure the cropped file is downloaded where applicable.