Use a log collector service to capturing logs from tests

[denis-tingaikin]

Motivation

Currently, we are going to use #244 for capturing logs from containers in k8s.

This solution is not perfect because potentially this increases CI time and it is not resistant to changing tests.

Solution

Use some kind of logs collecting service as https://www.fluentd.org/architecture

[Bolodya1997]

We have multiple issues regarding to the current logs capturing strategy:

1. We are not capturing logs for the test cleanup - if Client fails to Close the Connection for some reason, it affects subsequent tests, but there is nothing in logs for the initial failure.
2. We are not capturing logs for the previous-previous containers - if container restarts more than 1 time during the test, we will have logs only for the current and for the previous container run.
3. We are not capturing logs for the pods removed during the test - it affects all heal tests because we cannot get logs from the removed pod.

It doesn't look like fixing this from our side will be both working and good-looking, so +1 for using third-party tool for capturing logs.

[tiswanso]

We often use kubectl's cluster-info dump for this same purpose in CI. https://kubernetes.io/docs/reference/generated/kubectl/kubectl-commands#cluster-info

NOTE: prior to 1.19 its support of the namespace arg doesn't actually work correctly from what I've seen.