Ok, what I would like to know now is if adding a child will do automatic picking like containers do?
No, that's what containers are for.
Also, containers can only do that automatic picking because of the restriction around creating and destroying the container together. This enforces that there are always exactly the same number of instances of each object in the container, which is what allows the engine to be able to pick them all at once. If we made scene graph do that, we'd also need to reimpose the same restriction - and that would remove the ability to use scene graph dynamically (e.g. attaching and detaching different objects throughout the lifetime of one object). I think it's important to allow the dynamic usage of scene graph.
Besides, I'd say that well-designed software should not duplicate features in different places. If you want simultaneous picking, use a container. If you want objects to attach to each other, use scene graph. If you want both, use both!