refactor: Use non-zero initial error when the parameter value is zero for imporved performance and resilience

Author: AdrianDBS

minuit2.net/ParameterConfigurationExtensions.cs

@@ -11,7 +11,7 @@ public static MnUserParameterState AsState(this IEnumerable<ParameterConfigurati
 
     private static void Add(this MnUserParameterState state, ParameterConfiguration parameter)
     {
-        state.Add(parameter.Name, parameter.Value, parameter.Value * 0.01);
+        state.Add(parameter.Name, parameter.Value, parameter.Value == 0 ? 1 : parameter.Value * 0.01);
 
         if (parameter.IsFixed) 
             state.Fix(parameter.Name);

test/minuit2.net.UnitTests/The_simplex_minimizer.fails_for_the_following_challenging_problems.verified.txt

@@ -31,22 +31,6 @@
   BoxBodProblem.WithoutDerivatives (Fast),
   BoxBodProblem.WithoutDerivatives (Rigorous),
   BoxBodProblem.WithoutDerivatives (VeryRigorous),
-  FletcherPowellProblem.WithGradient (Balanced),
-  FletcherPowellProblem.WithGradient (Fast),
-  FletcherPowellProblem.WithGradient (Rigorous),
-  FletcherPowellProblem.WithGradient (VeryRigorous),
-  FletcherPowellProblem.WithGradientAndHessian (Balanced),
-  FletcherPowellProblem.WithGradientAndHessian (Fast),
-  FletcherPowellProblem.WithGradientAndHessian (Rigorous),
-  FletcherPowellProblem.WithGradientAndHessian (VeryRigorous),
-  FletcherPowellProblem.WithGradientHessianAndHessianDiagonal (Balanced),
-  FletcherPowellProblem.WithGradientHessianAndHessianDiagonal (Fast),
-  FletcherPowellProblem.WithGradientHessianAndHessianDiagonal (Rigorous),
-  FletcherPowellProblem.WithGradientHessianAndHessianDiagonal (VeryRigorous),
-  FletcherPowellProblem.WithoutDerivatives (Balanced),
-  FletcherPowellProblem.WithoutDerivatives (Fast),
-  FletcherPowellProblem.WithoutDerivatives (Rigorous),
-  FletcherPowellProblem.WithoutDerivatives (VeryRigorous),
   Mgh09Problem.WithGradient (Balanced),
   Mgh09Problem.WithGradient (Fast),
   Mgh09Problem.WithGradient (Rigorous),