Upgrade to DataInterpolations v7

[visr]

Instead of adding extra dummy data points to the arrays, we can now directly specify the extrapolation behavior in the DataInterpolations API:

https://docs.sciml.ai/DataInterpolations/stable/extrapolation_methods/

Fixes #1621.

[visr]

Oh this still fails a few tests, might need some work in DataInterpolations:

ERROR: LoadError: UndefVarError: `extrapolate_integral_left` not defined in `DataInterpolations`
Suggestion: check for spelling errors or missing imports.
Stacktrace:
  [1] integral(A::DataInterpolations.LinearInterpolation{Vector{Float64}, Vector{Float64}, Vector{Float64}, Vector{Float64}, Float64, (1,)}, t1::Float64, t2::Float64)
    @ DataInterpolations a:\.julia\packages\DataInterpolations\5Uosd\src\integrals.jl:27
  [2] formulate_storage!(current_storage::Vector{Float64}, tprev::Float64, t::Float64, flow_boundary::Ribasim.FlowBoundary{@NamedTuple{node_id::Vector{Int32}, time::Vector{Dates.DateTime}, substance::Vector{String}, concentration::Vector{Float64}}})
    @ Ribasim A:\repo\ribasim\Ribasim\core\src\solve.jl:182

[visr]

The Int() calls in this PR can be removed if we update to v7.1, SciML/DataInterpolations.jl#379.

[SouthEndMusic]

Oh this still fails a few tests, might need some work in DataInterpolations:
SciML/DataInterpolations.jl#386

[SouthEndMusic]

I gave it a try with the current master branch of DataInterpolations, and it still breaks, although differently than what @visr mentioned above. Now it (still) breaks on integrating an interpolation with cache_parameters = true where the upper integration bound coincides with the first data point.

SciML/DataInterpolations.jl#390

[SouthEndMusic]

Yep, with this fix nothings breaks anymore, the only failing tests are due to some slightly different simulation results.