v1.7

• implement occlusion culling, reduce draw calls but introduce new compute overhead
  • vertex shader don't have to process occluded objects
  • current implementation does not improve FPS but open new possibility to do so
• rework Camera API, Camera now responsible for defining viewport, render features, visibility
  • eliminate 2 concepts: render target and visibility task. now user only think about setting up camera
  • spot light now cost 2 depth map textures (2MB). point light now cost 6 depth map textures (6MB). they used to cost 7 depth map textures (7MB) regardless of light type