Null collision grid incorrect when ngrid not a factor of grid size

[WBMol]

The ngrid_<x,y,z> setting is currently such that if n is not a factor of the amount of grid cells in the given direction (x, y, z), the optical properties of the ngrid are not calculated accurately. This only shows up in specific cases with large gradients, such as stratus of a few grid cells deep.

Below an example.

For both simulations:
ktot = 64.

For exp2_100m, ngrid_z = 60 (results in miscalculation)
For exp2_100m_hg, ngrid_z = 32 (correct calculation).