Sorting of optimization result with priors involved doesn't work

[LarsFroehling]

If priors are included and the optimization fails nan is written in result.optimize_result for fval instead of the usual inf and therefore the sorting of result.optimize_result doesn't work properly. I am using the current develop branch of pyPESTO and AMICI. Can be reproduced e.g. with the Elowitz-model from the benchmark collection after adding normally distributed priors around zero for the initial concentrations which will surely make some runs fails. i.e.the parameter table becomes

parameterId	...	objectivePriorType	objectivePriorParameters
KM	...
background	...
eff	...
eff_GFP	...
init_GFP	...	normal	0; 1
init_GFP_mRNA	...	normal	0; 1
init_X_mRNA	...	normal	0; 1
init_X_protein	...	normal	0; 1
init_Y_mRNA	...	normal	0; 1
init_Y_protein	...	normal	0; 1
init_Z_mRNA	...	normal	0; 1
init_Z_protein	...	normal	0; 1
n_Hill	...
scale	...
sigma	...
tau_mRNA	...
tau_mRNA_GFP	...
tau_prot	...
tau_prot_GFP	...
tps_active	...
tps_repr	...

This altered file is attached.

Then

import pypesto
import pypesto.petab
import petab

petab_problem = petab.Problem.from_yaml("Benchmark-Models/Elowitz_Nature2000/Elowitz_Nature2000.yaml")
importer = pypesto.petab.PetabImporter(petab_problem)
model = importer.create_model(verbose=False)
obj = importer.create_objective()
problem = importer.create_problem(obj)
optimizer = pypesto.optimize.ScipyOptimizer()
result = pypesto.optimize.minimize(problem=problem, optimizer=optimizer,
                                   n_starts=20)

print(result.optimize_result.get_for_key('fval'))

outputs

[nan, -5.273827589103882, 14.223720396071306, 374502.751049893, nan, nan, nan, nan, nan, nan, 14.223697112161812, 157198.54584629158, 856908.663362333, nan, nan, nan, nan, -3.546770714032929, nan, 15242133642.081184]

parameters_Elowitz_Nature2000.zip