Improve potential name collision handling on taxonomy fixtures

`title` has a unique constraint in our DB, so we're relying on the
randomness of Faker's sentence generator not to create duplicates.
With `django_get_or_create` though, we can avoid the problem by not
creating a new row if a duplicate name happens to be generated.

directory/tests/factories/taxonomy.py

@@ -3,20 +3,23 @@
 from directory.models import Language, Country, Topic
 
 
-class LanguageFactory(factory.Factory):
+class LanguageFactory(factory.django.DjangoModelFactory):
     class Meta:
         model = Language
+        django_get_or_create = ('title',)
     # there were not suffient random words in faker, so we're using sentences
     title = factory.Faker('sentence', nb_words=3)
 
 
-class CountryFactory(factory.Factory):
+class CountryFactory(factory.django.DjangoModelFactory):
     class Meta:
         model = Country
+        django_get_or_create = ('title',)
     title = factory.Faker('sentence', nb_words=3)
 
 
-class TopicFactory(factory.Factory):
+class TopicFactory(factory.django.DjangoModelFactory):
     class Meta:
         model = Topic
+        django_get_or_create = ('title',)
     title = factory.Faker('sentence', nb_words=3)