Add ability to extract model parameters for simulation

[sebapersson]

See #85

The new function can cover pre-eq (by unfortunately having to rerun the pre-eq simulation but there is no way around), and model with condition specific parameters. Moreover, either a map or vector can be returned via retmap.

The function is tested.

I am struggling a bit with finding a good name. Feedback appreciated @TorkelE.

When all looks good I can improve docstring and merge with main.

[TorkelE]

Maybe something like get_model_fit or get fitted_model. get_fitted_u0_ps is descriptive (but not very pretty)

Would it be possible to make two sub functions, appending _ps and _u0 (extracting just those parts of the fit.

[sebapersson]

You are thinking two functions:

get_fitted_ps
get_fitted_u0

I like this solution the best as what the functions do is clear, and then the user will not have to confused whether the function returns u0 or p as first argument. It will also be easy to code.

If you do not have any objection I will go with this.

We can probably also provide the user with a function named get_fitted_ode (or similar) returning an ODE problem with ps and u0 correctly placed.

[TorkelE]

Sounds good. Actually, the get_fitted_ode returning the full ODE Problem sounds like a really good idea. That should totally get implemented at some point.

[sebapersson]

I splited into two functions, and added docstrings.

If you think it looks good I will merge with main.

It will also now be easy to do get_fitted_ode (I think), as it should simply be:

function get_fitted_ode(res::Union{PEtabOptimisationResult, PEtabMultistartOptimisationResult}, 
                                      petab_problem::PEtabODEProblem,
                                      condition_id::Union{String, Symbol}; 
                                      pre_eq_id::Union{String, Symbol, Nothing}=nothing)::ODEProblem

    u0, p = _get_fitted_parameters(res, petab_problem, condition_id, pre_eq_id, retmap)
    return remake(petab_problem.ode_problem, p=p, u0=u0)
end

[TorkelE]

Looks good.

Maybe, given that condition_id can be optional (if only 1 condition is used), I'd consider making it an arg, with a default value equal to the default condition name.

[sebapersson]

I agree, I have now added this. To also handle models with a single specified simulation condition if condition_id=nothing it now defaults to the first simulation condition.