dnsreplay had odd race conditions in sending/reseinding packets

dnsreplay had odd race conditions in sending/reseinding packets
We had logic to pre-read the first packet for better error reporting, and then not read it at the top of the while loop,
but process the pre-read packet. However, this logic kept engaging, causing us to retransmit the previous packet a lot, depending
on the size of our retransmit window, which in turn depended on the receiver characteristic. Horrible.
dnsreplay may previously have overstated the number of duplicate packets because of this.