Add BGC to ODE wrapper function

[radka-j]

This PR adds a bgc_to_ode wrapper function which can be used to turn a BGC model into an ODEProblem object. The differential equations and inference examples have been simplified accordingly. I have run both examples to check the outputs remain the same and I have also added a small unit test.

NOTES: there were issues with compiling a dependancy of DifferentialEquations.jl (BoundaryValueDiffEq.jl) and in the process of debugging that it occurred to me that as long as we aim to only solve ODEs we should use the more lightweight package OrdinaryDiffEq.jl (which also does not have this dependancy). So I updated our dependencies accordingly.

[nanophyto]

NOTES: there were issues with compiling a dependancy of DifferentialEquations.jl (BoundaryValueDiffEq.jl) and in the process of debugging that it occurred to me that as long as we aim to only solve ODEs we should use the more lightweight package OrdinaryDiffEq.jl (which also does not have this dependancy). So I updated our dependencies accordingly.

For SODEs we would also need StochasticDiffEq.jl. However, looking at the dependencies, this package also does not rely on BoundaryValueDiffEq.jl

Of course SODE's are a different beast and will require a different PR/more discussion, so sticking to OrdinaryDiffEq.jl for now makes sense to me.

[radka-j]

Exactly, I think we should add dependencies as we need them.

[radka-j]

The current set up will not work if the parameters are vectors or matrices. In those cases - what are the parameters that one is actually likely to want to do inference on? Is it all of them or is there an obvious subset to target?