Spatially varying diffusion scale when LYA_MULTIPLE_SCATTERING=True

[jordanflitter]

When LYA_MULTIPLE_SCATTERING=True, in principle, the diffusion scale varies locally as it depends on the ionization field at the cell and its surrounding (and also on the over-density field, but more weakly). For simplicity, we now consider the global ionization value, which is a good approximation before the onset of reionization. While this approximation breaks when reionization kicks in, under strong X-ray heating and strong Lya coupling the value of the spin temperature (and Lyman alpha flux) becomes irrelevant. This should be examined in the future in the case of weak X-ray heating and Lya coupling during the epoch of reionization.

I'm giving this issue a low priority because:

1. In most realistic scenarios, X-ray heating and Lya coupling are strong during reionization.
2. Fixing this issue would complicate things dramatically because a spatially varying filter implies that the box cannot be filtered in Fourier space via the convolution theorem!