Document that ugettext_lazy() result can't be used with arbitrary Python code

[Mike Edmunds]

The documentation ​currently states that:

The result of a ugettext_lazy() call can be used wherever you would use a unicode string (an object with type unicode) in Python.

However, the result of ugettext_lazy() does not pass isinstance(lazy_obj, unicode). Because instance checks are commonly used to detect text strings, the lazy objects are incompatible with some Python standard library code and other popular Python packages (such as requests).

Proposed doc patch clarifies that ugettext_lazy() can be used with other Django code, but that it should generally be converted to text before passing to arbitrary, non-Django Python code.

Discussion and examples on django-dev: ​https://groups.google.com/forum/#!topic/django-developers/eFsGf1ZrG7c

[Mike Edmunds]

Doc change: ​https://github.com/django/django/pull/7430

(Aside: I did also explore changing ugettext_lazy() to test as an instance of text_type. It's not totally infeasible, but would probably cause at least as many problems as it would solve. ​https://github.com/medmunds/django/pull/1 if you're curious.)

[Tim Graham <timograham@…>]

In d3708aeb:

Fixed #27382 -- Doc'd that ugettext_lazy() should be converted to text for non-Django code.

[Tim Graham <timograham@…>]

In 59001ca:

[1.10.x] Fixed #27382 -- Doc'd that ugettext_lazy() should be converted to text for non-Django code.

Backport of d3708aeb26032911897e213a300c8b8ea5c5a345 from master