LowerTriangularMatrix in PrognVars, DiagnVars +restructured

[milankl]

No description provided.

[milankl]

@white-alistair This PR is a bit hectic, but I've been restructuring the PrognosticVariables and DiagnosticVariables such that one can now do something like

julia> using SpeedyWeather
julia> progn,diagn,model = initialize_speedy();
julia> progn.
layers     lmax        mmax        n_leapfrog  nlev        pres

so PrognosticVariables contains another struct with layers and pres (as it's just surface separated). Hence, access to the variables is straight forward like

julia> progn.layers[1].leapfrog[1].vor
32×32 LowerTriangularMatrix{ComplexF32}:
...

and it should be super easy to loop over all layers like (which can be a single loop very early on at every time step)

for layer in progn.layers
    func!(layer,...)
end

[milankl]

With LowerTriangularMatrix as the default struct for spectral coefficients looping in spectral domain is either as before

for m in 1:mmax+1
    for l in m:mmax+1
        vor[l,m]

or more convenient (and faster)

for lm in eachharmonic(vor)
    vor[lm]

which is equivalent to

for lm in 1:length(vor)
    vor[lm]

[white-alistair]

@white-alistair This PR is a bit hectic, but I've been restructuring the PrognosticVariables and DiagnosticVariables such that one can now do something like

julia> using SpeedyWeather
julia> progn,diagn,model = initialize_speedy();
julia> progn.
layers     lmax        mmax        n_leapfrog  nlev        pres

so PrognosticVariables contains another struct with layers and pres (as it's just surface separated). Hence, access to the variables is straight forward like

julia> progn.layers[1].leapfrog[1].vor
32×32 LowerTriangularMatrix{ComplexF32}:
...

and it should be super easy to loop over all layers like (which can be a single loop very early on at every time step)

for layer in progn.layers
    func!(layer,...)
end

@milankl looks good, nice one!

I see that you moved the parametrization vars into a file column_variables.jl 👍 On a related note, say there are intermediate quantities used in the calculation of some parametrization, would you put those arrays in the IntermediateVariables struct or the ParametrizationVariables struct?

Let me know if you want an actual review of this at some point.

[milankl]

@white-alistair I answered that question in #118, I suggest creating a ColumnVariables struct.