Conversion to C++ headers: cmath, cfloat etc.

[chipbarnaby]

C++ applications should #include <cxxxx> as opposed to the C compatibility <xxxx.h>. That is, c<cmath>, not <math.h>.

Branch cmath makes the math.h -> cmath change.

Confused use of both in CSE caused results changes when vendor #includes were altered. Prior to fixes, cnglob.h included <math.h>. If a vendor include internally used cmath, compilation changes could result, e.g. for pow(), because math.h and cmath declare different overloads.

All CSE uses have been changed to cmath.

Lingering used of math.h that should be investigated / fixed --

D:\cse\vendor\kiva\src\libkiva\Functions.hpp(11):#include <math.h>
D:\cse\vendor\kiva\src\libkiva\Mesher.hpp(11):#include <math.h> <<<<<< also #include <cmath>
D:\cse\vendor\penumbra\vendor\glfw\src\input.c(35):#include <math.h>
D:\cse\vendor\penumbra\vendor\glfw\src\monitor.c(33):#include <math.h>
D:\cse\vendor\penumbra\vendor\glfw\src\win32_joystick.c(33):#include <math.h>
D:\cse\vendor\penumbra\vendor\libtess2\Source\geom.c(36):#include <math.h>
D:\cse\vendor\penumbra\vendor\libtess2\Source\tess.c(40):#include <math.h>

Other C-compatibility headers should be replaced incrementally (one at a time).