check the compute_cell_mask_slices computation

[tigloo123]

Seems to be in a reverse pyramid order (0 slice is the smallest)
Discards the adjusted (0.5) values of the height_map due to the zero_level