Better LOD for labels

[nitaku]

LOD for labels should use the number of cells rather than depth

[nitaku]

This turned out to be more problematic than what was expected. We returned to depth-based LOD, but we are still unsure. This aspect has to be considered in the overall LOD design.

[nitaku]

Another way to implement LOD could be giving each zoom step a corresponding depth level in the tree (and maybe a specific distance from the map). A smooth transition to the next level could be good.