CL2 should collect logs from system and test pods

Context: https://github.com/kubernetes/kubernetes/issues/87324

Clusterloader should collect logs from system (e.g. CoreDNS) and test (e.g. Prometheus) pods (equivalent to kubectl logs ...). This will improve our ability to debug test failed due to problems with those pods. Currently, we have very little insight into why they fail.

/area clusterloader
/priority critical-urgent

Start by tracing Clusterloader's existing handling of system and test pods, using the linked Kubernetes issue and the stated kubectl logs behavior as context. Done means Clusterloader collects logs from pods such as CoreDNS and Prometheus so failures in those pods provide useful debugging output.

Written by the indexing model from the issue text.