Merge pull request #2868 from sgnr/avoid-narrowing-cast-in-flo-round

Avoid a narrowing cast in flo_round under MRB_INT64.

src/numeric.c

@@ -436,7 +436,7 @@ flo_round(mrb_state *mrb, mrb_value num)
 {
   double number, f;
   mrb_int ndigits = 0;
-  int i;
+  mrb_int i;
 
   mrb_get_args(mrb, "|i", &ndigits);
   number = mrb_float(num);
@@ -451,7 +451,7 @@ flo_round(mrb_state *mrb, mrb_value num)
   }
 
   f = 1.0;
-  i = abs(ndigits);
+  i = ndigits >= 0 ? ndigits : -ndigits;
   while  (--i >= 0)
     f = f*10.0;