[aws][nlb] follow up of issue: targets receiving traffic when unhealthy (phase 02)

Note: as described on the linked card, AWS needs to finish the rollout of NLB improvement, so this task should be started on 2022H2, preferred August.

Folow up, phase 2, of NLB investigation of targets taking longer period do stop receive traffic after unhealthy.

All the details is available here:
1. Original investigation: SPLAT-307

2. Documentation with RCA

Goals of this task:

• Measure the results of total time the unhealthy nodes is receiving traffic - can be in the sample app (created on the doc) or kube-apiserver (described on BZ)

• Talk with AWS NLB engineers[1] if the numbers are higher than 60 seconds in the region us-east-1 using 3+ zones.

 [1] Contact of AWS engineers:

• elb-cset@amazon.com : The issue, initially, was discussed with Felipe da Silva