Try mitigating the confusion by the error messages we get in a default setup when running without an agent.

[klihub]

Basically everybody gets confused by the (gRPC-)error messages one gets when running with the default setup but without an agent, usually concluding that something is fatally wrong and the setup is completely dysfunctional. We should try to fix this.

There are a few options, but none of them seems outright the obviously correct choice.

The simplest option would be to disable the agent interface by default and add an extra (commented out) systemd environment variable as an almost self-documented way to allow easily enabling it. However this might give the impression that we try to discourage the usage of the agent which in reality we don't want. Then again, this could be documented.

Another solution would require a bit more changes, making the agent ingress and egress interfaces more aware of each other. This would require changing the current setup logic of the agent interface(s) so that the egress one (IOW, the one where cri-resmgr is the client and establishes the gRPC connection) would be considered dormant/disabled until a connection on the ingress interface from the agent is established. A Handshake/Hello/Register message could easily be added to the ingress agent gRPC protocol if needed with the explicit purpose of detecting incoming connections and triggering connection establishment on the egress interface.