Include water flux energy term in jacobian approximation

[juliasloan25]

To improve the stability of our longruns, we want to include the dT_rhoe/dvartheta_l term in our jacobian approximation. We have found that, in the case of high water flux near saturation, the dpsi/dtheta term that this depends on becomes very large and unstable to step explicitly. To demonstrate that the water flux energy is the culprit, we tried removing this term from the explicit tendency and saw improved stability.

Closed by #952
Working towards resolving #912

Plots

energy from water flux included in explicit tendency, dt = 450s

This build was before the change, with NaNs appearing in Africa after ~1 month. Note that this build ran for 4 years whereas the runs below ran for 2 years, and the plotting was changed after this build, introducing blocky coastlines in the plots below but not this one.

energy from water flux removed from explicit tendency, dt = 450s

This build, after removing the energy from water flux term from the explicit tendency, shows no NaNs in Africa after 2 years.

energy from water flux removed from explicit tendency, dt = 900s

In fact, in this build we are even able to take a larger timestep (450s -> 900s) after removing this term from the explicit tendency, and still don't see the NaNs in Africa (though some new ones appear in the northeast US, which could be due to our phase change approximation):