Reorganize Iperf error reporting

[Kuba314]

Description

Improve iperf reporting and reorganize the code a bit.

Tests

J:8583736 - second recipeset contains debug logs.

Reviews

@jtluka

Closes: #201

[jtluka]

Overall looks fine. Please address my comments. Also, do few test runs on the internal failing iperf test cases where we would benefit from the added logging.

[jtluka]

Looks good.

[olichtne]

just some minor comments to clear up

[Kuba314]

Currently I can't test this properly because of #354

EDIT: Planning to test this PR.

[olichtne]

looks good now.

[olichtne]

Currently I can't test this properly because of #354

EDIT: Planning to test this PR.

still planning to do some tests or is that also done now?

[Kuba314]

@olichtne So... I tested this PR but I'm not sure what the behaviour should be exactly...

I SSHed into the client agent and tried sending INT via kill:

• INTing iperf process directly makes the IperfClient module FAIL and even if there's data, after all iterations are done, there's a ZeroDivisionError when trying to get time_aligned_results.
• INTing the lnst-agent process which starts iperf does nothing (because KeyboardInterrupt is caught and there's a pass followed by another .communicate() call). I tried changing pass to server.send_signal(signal.SIG_INT), which results in the same behaviour as the previous point -> iperf gets SIGINT, finishes, but for some reason we get ZeroDivisionError.
• INTing main lnst-agent process kills the agent (Caught signal 2 -> dying on agent, host1 hard disconnected on controller side)

[Kuba314]

I added back the signal sending from the KeyboardInterrupt handler to forward it to the forked lnst-agent process and the iperf process. Should be good now. Ready to be merged.