Don't dump the peer table when unable to find a replacement for a duplicate.

This is a normal condition.  Here's how it occurs:

A and B have a connection, Y, which is fully up.  Then one of them
opens connection X in the other direction.  Let us suppose B's ID is
lower than A's, so that B is the peer which will take the initiative
when killing duplicate connections.

A			B
  IAM--X-->   <--X--IAM
              IAM--X-->

Here, B notices the duplicate:

              <--Y--DUP

Now, the network may send the data queued on connection Y before the
data queued on connection X, and/or A may notice it first:

  <--Y--DUP

Here, A looks for a replacement for Y and doesn't find one, because it
doesn't yet know that X is connected to the same peer as Y.

However, this doesn't hurt anything, because X will take over Y's
duties as far as being a channel from A to B goes as soon as the IAM
arrives, and that taking over of those duties is what duplicate
replacement is all about.