Use the overridable get_default_farm when starting S3 imports

Closes #1727 (closed)

Fixing the expected failures on starting processes with farms could have allowed us to catch this. I added some new tests with mocks to check that the correct get_default_farm is called, and opened https://redmine.teklia.com/issues/7025 to deal with all the other expected failures.

arkindex/documents/tests/test_corpus.py

@@ -74,7 +74,6 @@ class TestCorpus(FixtureAPITestCase):
             mock_now.return_value = FAKE_NOW
             cls.corpus_hidden = Corpus.objects.create(name="C Hidden")
 
-    @expectedFailure
     def test_anon(self):
         # An anonymous user has only access to public
         with self.assertNumQueries(4):
@@ -225,7 +224,6 @@ class TestCorpus(FixtureAPITestCase):
         self.assertEqual(len(data), 13)
         self.assertSetEqual({corpus["top_level_type"] for corpus in data}, {None, "top_level"})
 
-    @expectedFailure
     def test_mixin(self):
         vol1 = Element.objects.get(name="Volume 1")
         vol2 = Element.objects.get(name="Volume 2")
@@ -345,7 +343,7 @@ class TestCorpus(FixtureAPITestCase):
             "description": self.corpus_public.description,
             "public": True,
             "indexable": False,
-            "rights": ["read", "write", "admin"],
+            "rights": ["read"],
             "created": DB_CREATED,
             "authorized_users": 1,
             "top_level_type": None,