Zenodo cache should also preserve a list of the repos downloaded

[dgarijo]

So we can access them in the future.
We have to make this part a little more robust, in case we want to regenerate the full somef descriptions without issuing queries to Zenodo.