This is generic logic to check that all the server processes have
started, which seems useful to any consumer trying to start multiple
servers in the background. Therefore it makes more sense for it to
live in the serve module than in wptrunner.

We previously imported the serve module to build the config, but then
actually ran the server using `wpt serve` as a subprocess. This meant
we had to re-implement the logic to check the servers were started and
got it wrong (in particular, we were checking the http server but
using the https server for tests).

I removed the start() method and just had the constructor do the
connection because using the class in an unconnected state doesn't
make sense.