ENH: Run itkSimpleImageRegistrationTest fully in Debug builds

The test reduced the affine and deformable optimizers to a single
iteration when NDEBUG was undefined, a 2013 stopgap for Debug wall
time. A one-iteration registration stops at near-identity and tests
nothing useful, and its output required a Debug-only baseline.

That baseline was itkSimpleImageRegistrationTest.3.nii, dropped as
stale when the baselines were regenerated for the
JointHistogramMutualInformation fix in PR #6569. Since then every
Debug build that runs these tests has failed them: corista.kitware
Ubuntu-22.04-gcc11.4-TBB-Debug, ryzenator.kitware
Windows11-VS2022-Static-DebugTBB, RogueResearch19
Mac27.x-AppleClang-dbg-arm64, and the SimpleITK itk-coverage nightly.
ITK's own Azure Debug job excludes the RUNS_LONG label, so PR CI
never ran them.

Using the full iteration counts makes the Debug result bit-identical
to Release, so the existing baselines cover both and no Debug-only
baseline can go stale again. Measured on gcc 13.3 x86_64: the pair
takes 44 s wall under ctest -j2 (42 s + 44 s serial), against 5 s
for the one-iteration form.

Author: hjmjohnson

Modules/Registration/RegistrationMethodsv4/test/itkSimpleImageRegistrationTest.cxx

@@ -225,11 +225,7 @@ PerformSimpleImageRegistration(int argc, char * argv[])
   {
     itkGenericExceptionMacro("Error dynamic_cast failed");
   }
-#ifdef NDEBUG
   affineOptimizer->SetNumberOfIterations(std::stoi(argv[6]));
-#else
-  affineOptimizer->SetNumberOfIterations(1);
-#endif
   affineOptimizer->SetDoEstimateLearningRateOnce(false); // true by default
   affineOptimizer->SetDoEstimateLearningRateAtEachIteration(true);
   affineOptimizer->SetScalesEstimator(scalesEstimator1);
@@ -293,11 +289,7 @@ PerformSimpleImageRegistration(int argc, char * argv[])
 
   auto optimizer = GradientDescentOptimizerv4Type::New();
   optimizer->SetLearningRate(1.0);
-#ifdef NDEBUG
   optimizer->SetNumberOfIterations(std::stoi(argv[7]));
-#else
-  optimizer->SetNumberOfIterations(1);
-#endif
   optimizer->SetScalesEstimator(scalesEstimator);
   optimizer->SetDoEstimateLearningRateOnce(false); // true by default
   optimizer->SetDoEstimateLearningRateAtEachIteration(true);