debug conservation dependence on grid

[juliasloan25]

In #994, we found that the results of our conservation checks changed when the grid stretching used in atmos changed. We need to figure out why this is the case. See the thread in that PR for more details

[szy21]

We looked at the conservation tests together. The issue is likely because the coupler only updates exchange fluxes (e.g radiation and turbulent fluxes) at the end of each timestep, while in atmos they change at each stage. So there is a mismatch in the fluxes which results in errors in conservation. This doesn't matter for radiation because it is fixed for a certain time interval, so it affects water conservation more than energy conservation. Changing grid configuration may change the fluxes during the spin-up, which results in changes in the conservation error.