Revisit custom syncer logic

[mudler]

The syncer logic introduced in rancher-sandbox/rancheros-operator#22 relies on kubernetes logs to catch up third party custom syncers by using convoluted setup.

During the sync phase for the custom syncer, a pod is created with the container specified in the ManagedOSVersionChannel:

• The pod has a initcontainer with the container specified from the spec
• The initcontainer fetch the data somehow, and stores the json into a known location
• A container which takes the output json, displays it on console and sits and waits for the data to be captured from the operator via kubelet api logs, and gets cleaned up afterward

This card is a spike to find other solutions to it that doesn't involve logging.

One idea could be to stream back the JSON result directly into the operator via some service exposed in the cluster, where the operator can be reached.

Follow-up of rancher-sandbox/rancheros-operator#22