result: avoid dynamically-sized stack allocation

Otherwise, long field names have the potential to cause stack
overflow and perhaps slow down the MRI GC.  This reduces one
extraneous data copy while we're at it, too.

Using the checkstack.pl script in the Linux kernel reveals
no other dynamically sized stack allcations after this change.
usage: objdump -d mysql2.so | ~/linux/scripts/checkstack.pl x86_64

Author: Eric Wong

ext/mysql2/result.c

@@ -122,16 +122,12 @@ static VALUE rb_mysql_result_fetch_field(VALUE self, unsigned int idx, short int
 
     field = mysql_fetch_field_direct(wrapper->result, idx);
     if (symbolize_keys) {
-      char buf[field->name_length+1];
-      memcpy(buf, field->name, field->name_length);
-      buf[field->name_length] = 0;
-
 #ifdef HAVE_RB_INTERN3
-      rb_field = rb_intern3(buf, field->name_length, rb_utf8_encoding());
+      rb_field = rb_intern3(field->name, field->name_length, rb_utf8_encoding());
       rb_field = ID2SYM(rb_field);
 #else
       VALUE colStr;
-      colStr = rb_str_new2(buf);
+      colStr = rb_str_new(field->name, field->name_length);
       rb_field = ID2SYM(rb_to_id(colStr));
 #endif
     } else {