SB/KK microphysics cleanup

[CasparJungbacker]

Changes

In preparation for aerosols, some cleanup of the warm microphysics.

• Split SB and KK schemes into separate modules, similar to how the different advection schemes are implemented. This has several benefits:
  • The implementations of different sources are much more readable, as the subroutines are shorter.
  • Prevents using SB parameters in the KK scheme and vice-versa.
  • More modular approach makes it easier to add new schemes.
• Added more field_r's. When using FP32, all computations should now actually be done in FP32.
• Some (opinionated) formatting.

Tests

Used the benchmnark/rico case for testing. Started from a 24 hour warm start, then simulated another hour. Here are some plots of the different qr tendencies:

I think it would be a good idea to have someone who knows the microphysics go over the parameters, since modmicrodata has become very messy over the years.

@fjansson anything else that should be done?

[fjansson]

The curves look good.

This, in erfint

erfymin = erf(abs(ymin))
erfymax = erf(abs(ymax))
erfymin = sign(erfymin, ymin)
erfymax = sign(erfymax, ymax)

can be simplified to

erfymin = erf(ymin)
erfymax = erf(ymax)

since erf is antisymmetric, right? I think the abs and sign are remains of the previous polynomial expansion which may not have been antisymmetric. But the intrinsic erf should be. Anyway, I think this is used only if l_lognormal is true, which is not the default.