Test to see we don't send INVALID_ID response to duplicated packet

east:

- jacob-two-two: causes a copy of each packet to be sent after a small
  delay

- replay-duplicates: as the responder, this should cause an immediate
  re-transmit of the last response

The actual number is less - while performing DH for instance is no
packet to send out - so duplicates get dropped.  XXX: I'm not sure
this explains everything.

west:

- replay-duplicates: for every one of the above, inject two copies

- XXX: some states don't get properly tested because, during DH,
  packets are simply dropped; replay-encrypted (injected after crypto
  is working) and replay-corrupt should be implemented in IKEv1 :-)

Howver, it still seems a few short.
