vsock/virtio: fix accept queue count leak on transport mismatch

In the Linux kernel, the following vulnerability has been resolved: vsock/virtio: fix accept queue count leak on transport mismatch virtio_transport_recv_listen() calls sk_acceptq_added() before vsock_assign_transport(). If vsock_assign_transport() fails or selects a different transport, the error path returns without calling sk_acceptq_removed(), permanently incrementing sk_ack_backlog. After approximately backlog+1 such failures, sk_acceptq_is_full() returns true, causing the listener to reject all new connections. Fix by moving sk_acceptq_added() to after the transport validation, matching the pattern used by vmci_transport and hyperv_transport.