internal: fix GO_AWAY deadlock (#2391)

internal: fix GO_AWAY deadlock

A deadlock can occur when a GO_AWAY is followed by a connection closure. This
happens because onClose needlessly closes the current ac.transport: if a
GO_AWAY already occured, and the transport was already reset, then the later
closure (of the original address) sets ac.transport - which is now healthy -
to nil.

The manifestation of this problem is that picker_wrapper spins forever trying
to use a READY connection whose ac.transport is nil.

clientconn.go

@@ -1066,9 +1066,6 @@ func (ac *addrConn) createTransport(backoffNum int, addr resolver.Address, copts
 		case <-skipReset: // The outer resetTransport loop will handle reconnection.
 			return
 		case <-allowedToReset: // We're in the clear to reset.
-			ac.mu.Lock()
-			ac.transport = nil
-			ac.mu.Unlock()
 			oneReset.Do(func() { ac.resetTransport(false) })
 		}
 	}