Currently the lister are partially tested, mostly through mocks (and not easily debugged when something goes wrong).

Sometimes, after a refactoring, they do not break because the coverage is not enough.

Now that a swh-listers fixture got introduced (D2077), we should at least add one integration test per lister for the basic listing use case.

Plan:

• all listers: 1 basic listing (checking the lister's output in the scheduling db)

Ideally, adding those would be best:

• all listers: 1 basic listing failure (e.g lister misconfiguration ; phabricator lister requires api.token set or otherwise it fails, etc...)
• all listers: 1 listing with filtering on already seen origins
• pagination listers: multiple page calls (npm does now)