Make thermo constructors use kwargs

[charleskawczynski]

Description

This is a bit of a long-term goal, but it would be nice if the thermo state constructors used keywords and we had a single interface for all formulations:

function PhaseEquil{FT}(
    param_set::PS;
    ρ::Union{FT,Nothing}=nothing,
    e_int::Union{FT,Nothing}=nothing,
    q_tot::Union{FT,Nothing}=nothing,
    T::Union{FT,Nothing}=nothing,
    θ_liq_ice::Union{FT,Nothing}=nothing,
    p::Union{FT,Nothing}=nothing,
    maxiter::Union{Int,Nothing}=nothing,
    tol::Union{FT,Nothing}=nothing,
    ::Type{sat_adjust_method} = NewtonsMethod,
    ) where {FT<:Real,PS, sat_adjust_method}

    @assert q_tot ≠ nothing
    _FT = FT
    if e_int ≠ nothing && ρ ≠ nothing

        @assert T === nothing
        @assert θ_liq_ice === nothing
        @assert p === nothing
        maxiter === nothing && (maxiter = 3)
        tol === nothing && (tol = _FT(1e-1))

        # TODO: Remove these safety nets, or at least add warnings
        q_tot = clamp(q_tot, _FT(0), _FT(1))
        T = sat_adjust(e_int, ρ, q_tot, maxiter, tol, param_set)

    elseif θ_liq_ice ≠ nothing && ρ ≠ nothing

        @assert e_int === nothing
        @assert p === nothing
        @assert T === nothing
        maxiter === nothing && (maxiter = 30)
        tol === nothing && (tol = _FT(1e-1))

        # TODO: expose which numerical method to use for sat adjustment?
        T = saturation_adjustment_q_tot_θ_liq_ice(θ_liq_ice, ρ, q_tot, maxiter, tol, param_set)
        q_pt = PhasePartition_equil(T, ρ, q_tot, param_set)
        e_int = internal_energy(T, q_pt, param_set)

    elseif T ≠ nothing && p ≠ nothing

        @assert θ_liq_ice === nothing
        @assert e_int === nothing
        @assert ρ === nothing
        @assert maxiter === nothing
        @assert tol === nothing

        ρ = air_density(T, p, PhasePartition(q_tot), param_set)
        q = PhasePartition_equil(T, ρ, q_tot, param_set)
        e_int = internal_energy(T, q, param_set)

    elseif θ_liq_ice ≠ nothing && p ≠ nothing

        @assert T === nothing
        @assert e_int === nothing
        @assert ρ === nothing
        maxiter === nothing && (maxiter = 30)
        tol === nothing && (tol = _FT(1e-1))

        # TODO: expose which numerical method to use for sat adjustment?
        T = saturation_adjustment_q_tot_θ_liq_ice_given_pressure(θ_liq_ice, p, q_tot, maxiter, tol, param_set)
        ρ = air_density(T, p, PhasePartition(q_tot), param_set)
        q = PhasePartition_equil(T, ρ, q_tot, param_set)
        e_int = internal_energy(T, q, param_set)
        q_tot = q.tot

    else
        throw(ArgumentError("PhaseEquil kwarg combination incorrect."))
    end
    return PhaseEquil{_FT,PS}(param_set, e_int, ρ, q_tot, T)
end

Some things have changed since this was first written, but the idea should be clear.