Replace famous' internal coordinate storage with 4x4 matrices

[CompSciFutures]

The current coordinate system based on 3 vectors (position/scale/rotate) is confusing, limits the capabilities of the framework and seems to make a few things more difficult than they need to be internally.

Lets try to replace the current approach with a pure 4x4 transformation matrix internally so its more like a classical render tree, then maintain API support using a mixin.