Preparations to run inertial river routing on GPU

[BSchilperoort]

Issue addressed

Related to #646 - allows for parametric types of vectors in structs. The parametric types are also <:AbstractArray to allow for moving the data to GPU arrays instead of Julia native arrays.

To do;

• Modify function types to allow for <:AbstractFloat instead of Float64
• Refactor river network structs to allow for moving to GPU (related to Node to source edge mapping has ragged array; not GPU compatible #648)
• Refactor structs so that modules cannot be Nothing (see Skipping modules with !isnothing(...) incompatible with GPU kernels #647)
• Allow for configuration of GPU backend through config
• Implement GPU Kernel for river routing

Explanation

Adapt has been added as a dependency, to allow for moving arrays to GPU (e.g. CuArray or ROCArray)

Checklist

• Updated tests or added new tests
• Branch is up to date with master
• Tests & pre-commit hooks pass
• Updated documentation if needed
• Updated changelog.qmd if needed

[BSchilperoort]

Not sure what the {N} parameter was supposed to do here. Adding parametric types here broke that feature.

It also did not seem to be used anywhere, except for in the profile = FloodPlainProfile{n_depths}(; ... definition below. Removing this did not break any tests.

[BSchilperoort]

Tests were passing a Vector{Int}, which did not fail before despite the type of area being defined as Vector{Float64}. I guess some magical type casting was performed automatically?

Defining the type of area as <:AbstractArray{<:AbstractFloat} broke this however, so the arrays in this test had to be modified to have float elements.