Reactant primitive models

[maximilian-gelbrecht]

I don't want one monster PR, so I split the Reactant work a little.

#970 is about all the basics and getting the barotropic model going

Here, I continue with the primitive wet model

[maximilian-gelbrecht]

One thing that's annoying to deal with for Reactant compat is the initialize! of components that in the initialize construct some helper struct and then compute something with it. That's the case for e.g. orography with the interpolation, but also the land bucket model that first constructs a seasonal climatology land model.

I don't have the best way to deal with that figured out yet without going directly into the initialize! of these components and selectively apply a @maybe_jit

[milankl]

This is good to know for the parameterizations. For a lot of the surface fluxes we use condition ? a : b which can be easily written as ifelse but I'll try to directly implement this also for the convection scheme which has way more conditionals that I'll try to revisit in #976

[maximilian-gelbrecht]

Branches are not completely forbidding, we can @trace them in Reactant as well. But the binary search algorithm here was problematic with its data dependent while loop.