Add consistent environmental buoyancy gradients using mean fields and quadratures.

[ilopezgp]

This PR modifies the buoyancy_gradients closure and the struct used to compute them in the following ways:

• The outputs of buoyancy_gradients are modified to be the buoyancy gradients in the environment, and its dry and cloudy parts. Before, a couple of partial derivatives were outputted that were not used by the model, or just added up in functions in the code. I output the dry and moist derivatives instead because I may use them to simplify the mixing length formulation in the near future.
• Two implementations of the environmental buoyancy gradient are given. The first one corrects the implementation in Tan et al. (2018) in several ways. First, it exclusively uses environmental mean quantities for computation of buoyancy fluxes, which is consistent with not having access to a probabilistic cloud partition function. This is consistent with the original paper, although we later modified it in the code to use partial information from the quadratures. Second, it corrects the computation of the buoyancy gradient in the dry part from the one given in the paper. The derivation of this term in Tan et al (2018) is taken from Deardorff 1976 without consideration for the different assumptions allowed, and it is actually incorrect because θ_liq_ice != θ in the environment, just in the dry part of the environment. Third, it corrects the latent heat used to include effects of all condensate species.
• The second implementation is consistent with having Quadratures in the model. The derivatives are the same as in the previous case, therefore you will find that the buoyancy_gradients function is shared. The main distinction is in the arguments passed with the EnvBuoyGrad struct. In this case, we make full use of the quadrature information to compute the gradients. Safeguards are included for gradients computed in vanishing cloudy or dry regions. In these regions, we resort to the environmental mean gradients.

[ilopezgp]

For the MSE, I will wait until we decide whether we want to keep quadratures or mean as the default mode for the buoyancy flux. I prefer quadratures, since it should be more accurate.

[trontrytel]

Looks great to me!

My only comments are about the notation. I would vote for:

• using some abbreviation of saturated/unsaturated instead of cloud/dry
• agreeing on some convention about the thetas so we don't confuse ourselves later. Maybe something like \theta, \theta_v, \theta_liq_ice

But if the others disagree, we can keep the notation we used. Don't want to block this PR

[yairchn]

LGTM

[ilopezgp]

bors r+

[bors]

Build succeeded:

• buildkite/turbulenceconvection-ci
• docbuild
• format