14 update or create fix

[stefan-cardnell-rh]

For kafka_skip to work properly, We need to treat the following cases differently:

1. When we create Model instances directly, either with or without PK specified. e.g. User.objects.create(id=1, kafka_skip=True or User.objects.create(kafka_skip=True).

   In this scenario we never need to set the kafka_skip value since the kafka_skip value was either set directly or it will default to False since the model field has default=False.

2. When we retrieve an instance from DB, and save it again. E.g:

   instance = User.objects.create(id=1, kafka_skip=True)
   retrieved = User.objects.get(id=instance.id)
   retrieved.save()

   In this scenario, the DB loaded value (retrieved) has kafka_skip=True. Upon saving we want to reset kafka_skip to False only if kafka_skip was not set directly between get() and save().

Both scenarios are covered by adding a reset_kafka_skip variable and setting it correctly in _setattr_, __init__, and from_db.

Advantages:

• This works for all ways of creating/retrieving instances, including QuerySet.create, QuerySet.get_or_create, QuerySet.update_or_create.
• This works even when update_fields is not specified. Which is ideal as an open source package, since not everyone utilises update_fields.