Statistics improvements when omitted is used

[davidBar-On]

• Version of iperf3 (or development branch, such as master or
  3.1-STABLE) to which this pull request applies:
  master

• Issues fixed (if any):
  lost_percent greater than 100% #1489

• Brief description of code changes (suitable for use as a commit message):

Suggested fixes so Sender statistics when --omitt is used:

1. Fix issue lost_percent greater than 100% #1489: issue was caused because the sender statistics did not ignore the lost packets during the omit period.
2. Receiver statistics on sender side: use both packet count and omitted counts from receiver. Because of UDP packets delay, the number of ignored packets on sender and receiver are different - more packets are ignored on the sender since some of them did not arrive yet to the receiver at the end of the omit period. Therefore, the %lost packets that used the sender's ignored cont and the receiver's lost count was wrong.

[bmah888]

Nice diagnosis of the bug, and thanks for the PR!

There's an issue with interoperability with older versions of iperf3, see feedback comment. I'm not sure how to fix this yet but hopefully you have an idea since you've been thinking about this. Thanks!

[bmah888]

How to test this:

Server: iperf3 --server
Client: iperf3 --client client.example.org --time 5 --udp --bandwidth=1g (pick some bandwidth that's high enough to cause some packet loss, get a rough baseline of loss rate)
iperf --client client.example.org --time 5 --omit 3 --udp --bandwidth=1g (loss rate will probably be inflated before the fix in this PR, but should be similar with the PR)

Test with various combinations of old and new client and server. Also try --reverse mode.

Note that with a new client sending to an old server with --omit, the client will not be able to display the loss statistics from the server.