I can't answer on behalf of the developer's team, but I don't think it is *so* elementar. Buttons (and other GUI objects as well) are made of graphical elements (shape, borders, etc), fonts (text displayed), and event-listening methods (is the mouse over it? has the user clicked the object?), etc.

Thus, there are many different ways to create a button, depending on the display and behavior you expect, whereas there's only one way to display an image or print text. So, instead of using a built-in core library which may not fit their needs, users have the choice between many external librairies, and may choose their favorite.