Fix undesired reordering in queueing logic

[antoninbas]

C++ clocks can return the same value in consecutive calls, which was
causing potential reordering in the queueing logic, in particular for
rate-limited queues when the pps rate was set to 0.

This was causing some unit tests to fail.

Fixes #900

[codecov-commenter]

Codecov Report

Merging #903 into master will increase coverage by 0.00%.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master     #903   +/-   ##
=======================================
  Coverage   74.31%   74.32%           
=======================================
  Files         115      115           
  Lines       10243    10243           
=======================================
+ Hits         7612     7613    +1     
+ Misses       2631     2630    -1     

Impacted Files	Coverage Δ
src/bm_sim/learning.cpp	82.06% <0.00%> (+0.38%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 00b8e4a...605e788. Read the comment docs.

[jafingerhut]

Looks good to me. Nice catch figuring out what was going wrong here. I guess there is no practical need to deal with wrapping in time values or wrapping_counter values in BMv2, since we never expect it to run that long? You can imagine how much fun hardware folks have dealing with small values that do wrap. :-)

[antoninbas]

Even if the counter wraps, it may just create packet re-ordering for a very small time window, which is still strictly better than the current implementation. BTW even now, packet re-ordering is very uncommon. It seems to happen more frequently on some machines or depending on how bmv2 is run (e.g. in a docker container).