BUG: Fix divide-by-zero and integer truncation in CalculateFitError

Two bugs in CumulativeGaussianCostFunction::CalculateFitError():

1. Division by zero: if m_OriginalDataArray has zero elements, dividing
   by numberOfElements is undefined behavior. Guard added to the existing
   size-mismatch early return.

2. Integer truncation: `(1 / numberOfElements) * fitError` performs
   integer division, yielding 0 for any numberOfElements > 1 and making
   the RMSE always return 0.0. The correct RMSE formula is
   sqrt(fitError / numberOfElements).

Detected by clang-22 scan-build (core.DivideZero checker).

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Author: hjmjohnson

Modules/Numerics/Optimizers/src/itkCumulativeGaussianCostFunction.cxx

@@ -43,7 +43,7 @@ CumulativeGaussianCostFunction::CalculateFitError(MeasureType * setTestArray)
   // Use root mean square error as a measure of fit quality.
   const unsigned int numberOfElements = m_OriginalDataArray.GetNumberOfElements();
 
-  if (numberOfElements != setTestArray->GetNumberOfElements())
+  if (numberOfElements != setTestArray->GetNumberOfElements() || numberOfElements == 0)
   {
     return 1;
   }
@@ -52,7 +52,7 @@ CumulativeGaussianCostFunction::CalculateFitError(MeasureType * setTestArray)
   {
     fitError += Math::sqr(setTestArray->get(i) - m_OriginalDataArray.get(i));
   }
-  return (std::sqrt((1 / numberOfElements) * fitError));
+  return std::sqrt(fitError / static_cast<double>(numberOfElements));
 }
 
 double