[PATCH] Rename arguments in include/cmath to work around Newlib macro implementation using these particular names

[Eric Fiselier]

LGTM. I think we should change `__z` but I realize that if it doesn't cause any problems now it probably never will.

================
Comment at: include/cmath:873-874
@@ -872,4 +872,4 @@
 #if !(defined(_LIBCPP_MSVCRT) || defined(_AIX))
-inline _LIBCPP_INLINE_VISIBILITY float       frexp(float __x, int* __e) _NOEXCEPT       {return frexpf(__x, __e);}
-inline _LIBCPP_INLINE_VISIBILITY long double frexp(long double __x, int* __e) _NOEXCEPT {return frexpl(__x, __e);}
+inline _LIBCPP_INLINE_VISIBILITY float       frexp(float __lcpp_x, int* __e) _NOEXCEPT       {return frexpf(__lcpp_x, __e);}
+inline _LIBCPP_INLINE_VISIBILITY long double frexp(long double __lcpp_x, int* __e) _NOEXCEPT {return frexpl(__lcpp_x, __e);}
 #endif
----------------
Take it or leave it but maybe we should change `__e` as well. 

================
Comment at: include/cmath:1213-1215
@@ -1212,5 +1212,5 @@
 using ::fma;
 
-inline _LIBCPP_INLINE_VISIBILITY float       fma(float __x, float __y, float __z) _NOEXCEPT                   {return fmaf(__x, __y, __z);}
-inline _LIBCPP_INLINE_VISIBILITY long double fma(long double __x, long double __y, long double __z) _NOEXCEPT {return fmal(__x, __y, __z);}
+inline _LIBCPP_INLINE_VISIBILITY float       fma(float __lcpp_x, float __lcpp_y, float __z) _NOEXCEPT                   {return fmaf(__lcpp_x, __lcpp_y, __z);}
+inline _LIBCPP_INLINE_VISIBILITY long double fma(long double __lcpp_x, long double __lcpp_y, long double __z) _NOEXCEPT {return fmal(__lcpp_x, __lcpp_y, __z);}
 
----------------
Lets do it for `__z` just to be consistent.

http://reviews.llvm.org/D5080