use NAN and INFINITY

The macro str_to_mrb_float (strto[df]) converts a string to a number,
but these two macros can be used to directly get the special value.

The NAN macro requires quiet NaN support, but so does str_to_mrb_float.

This change also circumvents missing C99 support in Microsoft's C library.
Its (or ISO C90's) strtod doesn't parse "nan" or "inf".

src/numeric.c

@@ -304,8 +304,8 @@ flodivmod(mrb_state *mrb, mrb_float x, mrb_float y, mrb_float *divp, mrb_float *
   mrb_float mod;
 
   if (y == 0.0) {
-    div = str_to_mrb_float("inf");
-    mod = str_to_mrb_float("nan");
+    div = INFINITY;
+    mod = NAN;
   }
   else {
     mod = fmod(x, y);
@@ -775,7 +775,7 @@ fix_mod(mrb_state *mrb, mrb_value x)
     mrb_int mod;
 
     if (mrb_fixnum(y) == 0) {
-      return mrb_float_value(mrb, str_to_mrb_float("nan"));
+      return mrb_float_value(mrb, NAN);
     }
     fixdivmod(mrb, a, mrb_fixnum(y), 0, &mod);
     return mrb_fixnum_value(mod);
@@ -805,8 +805,8 @@ fix_divmod(mrb_state *mrb, mrb_value x)
     mrb_int div, mod;
 
     if (mrb_fixnum(y) == 0) {
-      return mrb_assoc_new(mrb, mrb_float_value(mrb, str_to_mrb_float("inf")),
-        mrb_float_value(mrb, str_to_mrb_float("nan")));
+      return mrb_assoc_new(mrb, mrb_float_value(mrb, INFINITY),
+        mrb_float_value(mrb, NAN));
     }
     fixdivmod(mrb, mrb_fixnum(x), mrb_fixnum(y), &div, &mod);
     return mrb_assoc_new(mrb, mrb_fixnum_value(div), mrb_fixnum_value(mod));