common: treat all "all-channels" errors as if they were warnings.

This is in line with the warnings draft, where all-zeroes in a
channel_id is no longer special (i.e. it will be ignored).

But gossipd would send these if it got upset with us, so it's best
practice to ignore them for now anyway.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Changelog-Added: Protocol: we treat error messages from peer which refer to "all channels" as warnings, not errors.

lightningd/peer_control.c

@@ -375,7 +375,7 @@ void channel_errmsg(struct channel *channel,
 		    struct per_peer_state *pps,
 		    const struct channel_id *channel_id UNUSED,
 		    const char *desc,
-		    bool soft_error,
+		    bool warning,
 		    const u8 *err_for_them)
 {
 	notify_disconnect(channel->peer->ld, &channel->peer->id);
@@ -388,13 +388,14 @@ void channel_errmsg(struct channel *channel,
 	}
 
 	/* Do we have an error to send? */
-	if (err_for_them && !channel->error)
+	if (err_for_them && !channel->error && !warning)
 		channel->error = tal_dup_talarr(channel, u8, err_for_them);
 
 	/* Other implementations chose to ignore errors early on.  Not
 	 * surprisingly, they now spew out spurious errors frequently,
-	 * and we would close the channel on them. */
-	if (soft_error) {
+	 * and we would close the channel on them.  We now support warnings
+	 * for this case. */
+	if (warning) {
 		channel_fail_reconnect_later(channel, "%s: (ignoring) %s",
 					     channel->owner->name, desc);
 		return;