normcdf is not supported using MSVC for use_dll

normcdf has a one line implementation using standard library functions (included in MSVC), namely:

double normcdf(double value)
{
   return 0.5 * erfc(-value * M_SQRT1_2);
}

It seems a little unnecessary for this not to be supported with MSVC.

Implementation taken from: https://stackoverflow.com/questions/2328258/cumulative-normal-distribution-function-in-c-c

@houtanb Could we get this into 4.5? We already have a check for MSVC like

#ifdef _MSC_VER
# define nearbyint(x) (fabs((x)-floor(x)) < fabs((x)-ceil(x)) ? floor(x) : ceil(x))
#endif

in the use_dll_static and use_dll_dynamic-files so adding this might be not that much work.

@houtanb While we are at it, we should try to lift the restriction on acosh, asinh, atanh. MSVC is supporting them since MSVC 2012, i.e. _MSC_VER bigger than 1600. See http://stackoverflow.com/questions/15539116/atanh-arc-hyperbolic-tangent-function-missing-in-ms-visual-c

allowing acosh, asinh, atanh, normcdf with MSVC >= 11.0 in 4255be00

@JohannesPfeifer @tholden I only find erfc support by msvc going back to 2012. Was this supported before? https://msdn.microsoft.com/en-us/library/hh308292(v=vs.110).aspx

@houtanb It seems I was wrong. erfc was introduced in MSVC together with the other functions like acosh. So d7d2e4fd was wrong and should be reverted, while https://github.com/DynareTeam/dynare/commit/4255be00b8a10226470c4cc1bc5a514ce37041bc was ok.

@JohannesPfeifer we can implement the solution here for MSVC < 2012

@houtanb We could. But given that MSVC 2015 is free and #1227 (closed) seems to mandate that version as well in master, we should probably not devote resources to this.

What is still missing to function correctly with normcdf is a

#ifdef _MSC_VER
#define _USE_MATH_DEFINES
#endif

Before

#include <math.h>

in the _dynamics.c and _static.c files to define the mathematical constants. See https://msdn.microsoft.com/library/4hwaceh6.aspx