Improved traversal approach for predictable 2d image formats displayed in 2d

Related to #943

WMTS, DZI, COG, etc are all 2d image formats with (presumably) consistently sized tiles which means you can very quickly determine which LoD and which sub tiles need to be rendered when the render perspective is limited to 2D. Ie if you have a specific 2d window and the tiles are all the same size, you can immediately determine which LoD level will need to be rendered for the tiles, and you can figure out how many tiles from 0, 0 the window is due to he consistent tile size. This also means the tile hierarchy does not need to be expanded into bounds higherarchy for traversal, as well.

When rendering in 3d these optimizations are not possible but if a user wants to display images in 2d it may be possible to add some improvements for traversing and rendering in these cases.

Start by reviewing the traversal and rendering paths for 2D image views, along with related issue #943. The work is complete when predictable tile formats such as WMTS, DZI, and COG can use direct LoD and tile selection without expanding the hierarchy into bounds for 2D rendering.