Bugfix: Huggingface dataset uploader

[josvandervelde]

Andrej found a bug in the HuggingFace uploader: as repo_id it used username/dataset_name_cleaned, while you'd expect the dataset_name to already include the username.

In this PR, this is fixed:

• Instead of using dataset.name as repo_id, the dataset.platform_resource_identifier is used, because this identifier should uniquely identify the dataset on HF.
• An error is thrown if you try to upload data to HF, if the platform of the mentioned dataset is not "huggingface" or the dataset.platform_resource_identifier is not a valid <namespace>/<dataset_name>.
• On POST/PUT the metadata of a resource, there are now checks:

• If platform=huggingface, the platform_resource_identifier should be a valid repo_id (unfortunately, this might be an id without <namespace>, because we have to support legacy HF datasets, and I couldn't think of a clean way to allow legacy datasets only for the connector, not for the POST/PUT routes, without slightly complicated and ugly code
• If platform=openml, the platform_resource_identifier should be a positive integer
• Other platforms are not checked yet. Jean Matias is currently busy with Zenodo, I'll ask if he could add a check in his PR

[andrejridzik]

I would split this check into two separate ones and provide different error message

[josvandervelde]

Done! 3f0e878

[josvandervelde]

As discussed:

It would probably be better to have an endpoint to which you can give all metadata + the data. Then it can

1. check the metadata (return if problems)
2. check if it's possible to create repo (return if problems)
3. create repo on HF
4. upload data (return if problems)
5. create AIoD metadata
   We'll leave that out-of-scope for this PR