Revisit the overlap assumption

[szy21]

Currently, we are using the maximum random overlapping assumption from RRTMG. (Note that it is different from the one in RRTMGP.)

Robert notes that maximum-random overlap has the disadvantage of being closely tied to the vertical discretization. Overlap formulations which rely on correlations with vary with physical distance, as described in the second paper below, are more robust. https://rmets.onlinelibrary.wiley.com/doi/abs/10.1256/qj.03.99 https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2004JD005100

Right now, there are no physical tests for the overlap assumption. Tapio suggested that we use 3D radiation as a reference (see Clare's paper).

The original RRTMG also has some useful tests: https://github.com/AER-RC/RRTMG_SW

[szy21]

cc @tapios