Modify autodiscovery for kubelet considering new Arch

[paologallinaharbur]

Taking into account that the process needs to be longrunning verify

• what we are actually caching/discovering regarding kubelet
• if we can simplify that code or improve performances thanks the re-arch

[invidian]

Some questions and thoughts arised while we were discussing #217:

• Kubelet discovery process is not about where to connect, almost, as most of the time we connect via host IP, however, in cases like Fargate, we will be connecting over API server proxy endpoint.
• Kubelet discovery is mainly about how to connect to kubelet endpoint, so either with TLS, without TLS or with insecure TLS.
• We need to decide on strategy for discovery while having a long-running process. Do we want to make a discovery only once during initialization of the integration and then just re-use this? Kubelet might be re-configured in the meanwhile, however, for such situations, the recommended way is to taint and drain the node, change the configuration and then schedule the workloads again. However, DaemonSets does not get unscheduled in such cases, so agent will not be restarted. But generally speaking, this seems like an exotic feature to support.