Store the result of MetadataFetcher.get_parent_origins

So it is available to concrete loaders.

The Git loader will use it to incrementally load new origins by fetching
the last snapshot of a parent, instead of starting from scratch.