I want to be able to ingest a big dataset directly from an archive like Zenodo without needing to download locally

When trying to reproduce [this paper](https://doi.org/10.5194/wes-10-435-2025), there are quite a few large data files on a university data archive. I would like to be able to add the large files to the project without downloading them. Their MD5 is known, so we could potentially add them to the DVC cache directly, then perhaps a remote machine can process the stage they're involved in.

Related to #308.

Does `dvc import` help here? Actually it looks like [`dvc import-url`](https://dvc.org/doc/command-reference/import-url) might be helpful.