Added a json_object_new_double_s() convenience function to allow an exact string representation of a double to be specified when creating the object and use it in json_tokener_parse_ex() so a re-serialized object more exactly matches the input.

Add json_object_free_userdata() and json_object_userdata_to_json_string() too.

tests/test_parse.expected

@@ -5,7 +5,7 @@ new_obj.to_string()="ABC"
 new_obj.to_string()=null
 new_obj.to_string()=true
 new_obj.to_string()=12
-new_obj.to_string()=12.300000
+new_obj.to_string()=12.3
 new_obj.to_string()=[ "\n" ]
 new_obj.to_string()=[ "\nabc\n" ]
 new_obj.to_string()=[ null ]