Tradition, 3d videogames started doing it like that because of how computers worked 40 years ago, then devs got used to think about 3d space that way and it stuck. Essentially videogames think about visual depth. And yes, the physics engines for videogames usually account for that and use their own transformations of formulas because they are rarely simulating anything more complex than rigid body physics. Advanced simulations aren't any harder for devs, all the transformations are abstracted away with libraries.

In the end they are just reference frames and up is whatever you want it to be. As Wikipedia puts it eloquently: “Unlike most mathematical concepts, the meaning of a right-handed coordinate system cannot be expressed in terms of any mathematical axioms. Rather, the definition depends on chiral phenomena in the physical world, for example the culturally transmitted meaning of right and left hands, a majority human population with dominant right hand, or certain phenomena involving the weak force.”