[Insight-developers] Be careful when using NumericTraits<>::min

[Lorensen, William E (Research)]

Folks,
	The C++ min() for floats and doubles behaves differently than it does for other types.
For all but float and double, min() returns what you might expect, the minimum value for the type.
For float and double, it returns the least significant value. itk has a NonpositiveMin which
consistently represents the minimum value for the type.

There were several incorrect uses of min() in itk. Most notable, the MinimumMaximumImageFilter.
Others occurred in statistics classes. I changed these to NonpositiveMin(). The FEM classes used
<double>min() correctly, so I left them alone.

Here is the output of a new test Common/itkNumericTraitsTest

itk::NumericTraits<char>
        min(): -128
        max(): 127
        NonpositiveMin(): -128
itk::NumericTraits<unsigned char>
        min(): 0
        max(): 255
        NonpositiveMin(): 0
itk::NumericTraits<short>
        min(): -32768
        max(): 32767
        NonpositiveMin(): -32768
itk::NumericTraits<unsigned short>
        min(): 0
        max(): 65535
        NonpositiveMin(): 0
itk::NumericTraits<int>
        min(): -2147483648
        max(): 2147483647
        NonpositiveMin(): -2147483648
itk::NumericTraits<unsigned int>
        min(): 0
        max(): 4294967295
        NonpositiveMin(): 0
itk::NumericTraits<long>
        min(): -2147483648
        max(): 2147483647
        NonpositiveMin(): -2147483648
itk::NumericTraits<unsigned long>
        min(): 0
        max(): 4294967295
        NonpositiveMin(): 0
itk::NumericTraits<float>
        min(): 1.17549e-38
        max(): 3.40282e+38
        NonpositiveMin(): -3.40282e+38
itk::NumericTraits<double>
        min(): 2.22507e-308
        max(): 1.79769e+308
        NonpositiveMin(): -1.79769e+308