Document FrameLocalsProxy behaviour quirks

Following PEP 667, https://docs.python.org/dev/reference/datamodel.html#frame-objects mentions that frame.f_locals may be a write-through proxy object, but doesn't really explain what that means.

Once https://github.com/python/peps/pull/3845 lands PEP 667 will contain such a description, but the main docs should at least cover the key behaviours:

• new instance on every attribute access
• two proxies with the same keys and values will still be unequal if they refer to different frames
• extra keys can be stored and are stored on the frame itself so they're shared across all proxies for the same frame
• keys corresponding to local variables can't be deleted
• .copy() (and any other API that produces a new container) returns a regular dict instance

Suggested location would be as a new 4th subsection under the frame objects section linked above. A "see also" from that new subsection back to PEP 667 would also be a useful addition.

Commencez par la section consacrée aux objets frame dans la documentation du modèle de données Python et examinez PEP 667, y compris la pull request du PEP mentionnée. Ajoutez une quatrième sous-section décrivant les cinq comportements listés de FrameLocalsProxy ainsi qu’un lien « voir aussi » vers PEP 667. Le travail est terminé lorsque la documentation principale explique l’identité du proxy, les clés supplémentaires partagées, les variables locales indélébiles et les API qui produisent des dictionnaires.