Changed console and filebackend to use msg.as_bytes to output the data as it would get send via smtp.

This is to provide a consistent interface (namely bytes) for the smtp
backend which after all sends bytes over the wire; encoding with as_string
yields different results since mails as unicode are not really specified.

as_string stays for backwardscompatibilty mostly and some debug outputs.
But keep in mind that the output doesn't match as_bytes!

Made it use 'AUTOINCREMENT' suffix for PK creation. This way it doeesn't
regress when compared with the 'traditional' DB backend creation
infrastructure.

Refs #10164.

Thanks to Arfrever for the report and Aymeric for the review.

Thanks kz26 for the report and Tim Graham for the excellent
review.

The last component of the dotted path to the application module is
consistently referenced as the application "label". For instance it's
AppConfig.label. appname could be confused with AppConfig.name, which is
the full dotted path.

register_model is called exactly once in the entire Django code base, at the
bottom of ModelBase.__new__:

    new_class._meta.apps.register_model(new_class._meta.app_label, new_class)

ModelBase.__new__ exits prematurely 120 lines earlier (sigh) if a model with
the same name is already registered:

    if new_class._meta.apps.get_registered_model(new_class._meta.app_label, name):
        return

(This isn't the exact code, but it's equivalent.)

apps.register_model and apps.get_registered_model are essentially a setter and
a getter for apps.all_models, and apps.register_model is the only setter. As a
consequence, new_class._meta.apps.all_models cannot change in-between.

Considering that name == new_class.__name__, we can conclude that
register_model(app_label, model) is always called with such arguments that
get_registered_model(app_label, model.__name__) returns None.

Considering that model._meta.model_name == model.__name__.lower(), and looking
at the implementation of register_model and get_registered_model, this proves
that self.all_models[app_label] doesn't contain model._meta.model_name in
register_model, allowing us to simplify the implementation.

Refs #21692 and 3efd1b8b.

Fixed #21692 -- Quote table name when creating it.

The u prefix looks bad on Python 2.

Thanks nedbatchelder and claudep.

Mostly for consistency with MODELS_MODULE_NAME; it's unlikely to change.

Thanks oubiga for the report.

An admin view performed the capitalization in the template, unlike all others.

Refs #10436.

Thanks Aymeric for the suggestion.

Thanks cjerdonek for the suggestion.

It raises ImportError whenever an entry in INSTALLED_APPS points
(directly or indirectly via AppConfig.name) to a non-existing module
and ImproperlyConfigured in all other cases.

Catching ImportError and re-raising ImproperlyConfigured tends to make
circular imports more difficult to diagnose.

Eliminated the app_ prefix that was more confusing than useful.

Previously the _apps/models_loaded flags could remain set to False if
set_installed_apps exited with an exception, which is going to happen as
soon as we add tests for invalid values of INSTALLED_APPS.

Thanks thepapermen for the report and Carl Meyer for the review.

Fixed a crash when executing changepassword command when the user object
representation contained non-ASCII characters.

Removed redundant line