Add computation of averages for subsidence forcing to Bomex example

[glwagner]

@mmr0 this PR shows how you can add a computation of the averages as a callback to complete the subsidence forcing.

[mmr0]

Fu_subsidence = Forcing(u_subsidence, discrete_form=true, parameters=(;u_avg=u_avg_f, wˢ))
Fv_subsidence = Forcing(v_subsidence, discrete_form=true, parameters=(;v_avg=v_avg_f, wˢ))
Fθ_subsidence = Forcing(θ_subsidence, discrete_form=true, parameters=(;θ_avg=θ_avg_f, wˢ))
Fq_subsidence = Forcing(q_subsidence, discrete_form=true, parameters=(;q_avg=q_avg_f, wˢ))

Found I needed to change the syntax of parameters (otherwise I got an error along the lines of
ERROR: LoadError: type Field has no field wˢ

Also q_avg=u_avg_f is a good way to make lots of clouds!! haha

[navidcy]

(u_avg=u_avg_f; wˢ)

seems typo; I think @glwagner was trying to construct a named tuple, e.g., exactly what you suggest above @mmr0

[navidcy]

does this PR supersedes #15?

[navidcy]

Looks good to me; @mmr0 can you have another look?

[navidcy]

This runs on GPU.

This is zeros:

julia> qˡt = FieldTimeSeries("bomex.jld2", "qˡ")
128×128×150×61 FieldTimeSeries{InMemory} located at (Center, Center, Center) of qˡ at bomex.jld2
├── grid: 128×128×150 RectilinearGrid{Float64, Periodic, Periodic, Bounded} on CPU with 3×3×3 halo
├── indices: (:, :, :)
├── time_indexing: Linear()
├── backend: InMemory()
├── path: bomex.jld2
├── name: qˡ
└── data: 134×134×156×61 OffsetArray(::Array{Float64, 4}, -2:131, -2:131, -2:153, 1:61) with eltype Float64 with indices -2:131×-2:131×-2:153×1:61
    └── max=0.0, min=0.0, mean=0.0

Is this part of the issue the PR is trying to address?

[glwagner]

Bomex should have cloud liquid. However, lack of cloud formation may be due to multiple factors. Therefore, this PR just implements one step towards a correct bomex. This PR does not aim to validate bomex. It should be merged if possible so that work can continue on validation.

[navidcy]

Let's merge then and continue!