Add minimum exchange coeff + exponentially decaying diffusion to Frierson functions

[akshaysridhar]

Closes #3228
summary_2D.pdf
summary.pdf

edt instantaneous at upper levels shows ~ 0 diffusivity.

[szy21]

I left some minor comments, but feel free to merge it if you don't agree with them. We can talk later. It would be good to try this with the longrun tonight.

[akshaysridhar]

@szy21 @trontrytel With the latest commit 1998e87

We get the following results, 30d avg at day ~360 (see pdf summary plots below) for the aquaplanet 1M microphysics test ( no topography )

summary_2D.pdf
summary.pdf

@szy21 Ri_c is now the default value; however mean surface windspeed at the 15m level appear to be ~7m/s so empirically this O(10) gust norm is perhaps reasonable (we can revisit this). With h_boundary_layer set to level 1 prior to the column-wise check, our runs blow up during the spinup period within hours of simulation startup presumably due to specific columns where we don't meet the Ri <= Ri_c criterion.

[szy21]

Sounds good. I still think h_boundary = 0.1 * domain_top is too high though, and may hide some problems. So the minimum diffusivity when z > h is insufficient to stabilize it. Could you try setting h_boundary to something like 500m and see if it helps?