veth: reduce stat overhead

veth stats are a bit bloated. There is no need to account transmit
and receive stats, since they are absolutely symmetric.

Also use a per device atomic64_t for the dropped counter, as it
should never be used in fast path.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

include/linux/netdevice.h

@@ -1284,6 +1284,7 @@ struct net_device {
 		struct pcpu_lstats __percpu	*lstats; /* loopback stats */
 		struct pcpu_tstats __percpu	*tstats; /* tunnel stats */
 		struct pcpu_dstats __percpu	*dstats; /* dummy stats */
+		struct pcpu_vstats __percpu	*vstats; /* veth stats */
 	};
 	/* GARP */
 	struct garp_port __rcu	*garp_port;