Currently, a failure happens for some reason. As some internal state variables
are not initialized (they are in `fetch_data` but not in `__init__`), this
makes the `get_snapshot` method fails because of it. Then that's making the
visit update step fails.
Adding the proper initialization for those variables in the `__init__` method
should fix it. That will allow the failed state of the visit to be updated
properly in the storage. Associating an empty snapshot to the failed visit.
This should fix issues like [1]
https://sentry.softwareheritage.org/share/issue/ace9143c0f6244b49b18cca49a5aa809/