Optimise the LowPoly Renderer

[mario-deluna]

The current implementation is really just for testing and getting something on the screen.

An optimised system should properly sort, group and batch rendering calls whenever possible.

• Implement static renderable as separate components.
• Organise renderables in Quadtree or octree to implement basic culling.
• Extract the renderables into render call like data structures
• Sort the "renderList" entities by shader, buffer, material Z etc..
• Batch static renderables together per sector.
• Test if PHP bottlenecks when frustum culling dynamic entities.