Receive loop is racey

[BrennanConroy]

TEST(websocket_transport_connect, transport_destroyed_even_if_disconnect_not_called)
On fast computers this fails very often because the transport shared pointer is acquired in the receive loop before the old receive loop finishes and destroys it acquired shared pointer. This could potentially "never" finish or take a very long time in production, wasting cpu cycles/resources.

[moozzyk]

The best way of fixing this might be to get rid of the receive loop altogether. The casablanca's websocket client is going to have 'message_received' and 'websocket_closed' events in the (hopefully) next version and we should look into using these instead of having our receive loop. It should also simplify the websocket transport.