channeld: don't crash if remote asks for unnecessary retransmit.

We assume they haven't acknowledged it, but they might have (seems
like a bug though).  We could hang up on them, but we'll happily
continue.

lightning_channeld: channeld/channel.c:1702: resend_commitment: Assertion `peer->revocations_received == peer->next_index[REMOTE] - 2' failed.
lightning_channeld: Fatal signal 6
0x559919c35dcd crashdump
        common/daemon.c:37
0x7f76d5be27ef ???
        ???:0
0x7f76d5be277f ???
        ???:0
0x7f76d5be4379 ???
        ???:0
0x7f76d5bdab46 ???
        ???:0
0x7f76d5bdabf1 ???
        ???:0
0x559919c29337 resend_commitment
        channeld/channel.c:1702
0x559919c297cf peer_reconnect
        channeld/channel.c:1853
0x559919c2b3d5 init_channel
        channeld/channel.c:2518
0x559919c2b6f8 main
        channeld/channel.c:2580
0x7f76d5bcd3f0 ???
        ???:0
0x559919c244d9 ???
        ???:0
0xffffffffffffffff ???
        ???:0

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

channeld/channel.c

@@ -1703,7 +1703,13 @@ static void resend_commitment(struct peer *peer, const struct changed_htlc *last
 	enqueue_peer_msg(peer, take(msg));
 	tal_free(commit_sigs);
 
-	assert(peer->revocations_received == peer->next_index[REMOTE] - 2);
+	/* If we have already received the revocation for the previous, the
+	 * other side shouldn't be asking for a retransmit! */
+	if (peer->revocations_received != peer->next_index[REMOTE] - 2)
+		status_unusual("Retransmitted commitment_signed %"PRIu64
+			       " but they already send revocation %"PRIu64"?",
+			       peer->next_index[REMOTE]-1,
+			       peer->revocations_received);
 }
 
 /* Our local wrapper around read_peer_msg */