2n-zn2 iavf: two-band structure in 1c mrr tests

[vrpolakatcisco]

This is interesting, because 2c tests use 4 workers. Looking at trending [0], 1c tests have similar stdev but no band structure. Comparing good [1] and bad [2] telemetry I do not see a good reason why RX nodes are able to read somewhat different number of packets.

[0] 2n-zn avf l2-patch [1] https://logs.fd.io/vex-yul-rot-jenkins-1/csit-vpp-perf-mrr-daily-master-2n-zn2/1223/log.html.gz#s1-s1-s1-s5-s11-t2-k2-k9-k10-k14-k1-k1-k1-k1 [2] https://logs.fd.io/vex-yul-rot-jenkins-1/csit-vpp-perf-mrr-daily-master-2n-zn2/1224/log.html.gz#s1-s1-s1-s5-s11-t2-k2-k9-k10-k14-k1-k1-k1-k1