On the Cost of Using Happy Eyeballs for Transport Protocol Selection