Fix gcp resource detection when using workload identity

[dashpole]

Description:

I was testing #10347, which updated gcp resource detection, and got this error on clusters with workload identity: failed to detect resource {"kind": "processor", "name": "resourcedetection", "pipeline": "logs", "error": "metadata: GCE metadata \"instance/name\" not defined"}. It turns out that on GKE, only instance/hostname is available, not instance/name. Hostname is an fqdn, not the name of the instance. E.g. gke-wi-collector-test-default-pool-5a7b28f7-nxwt.c.my-project.internal, so we shouldn't use that as the hostname. It would be nice to have the instance name, but that might need to be obtained through the downward API instead of the GKE metadata server...

Testing:

Fixed unit tests, and manually tested.

[pmm-sumo]

Should we perhaps just trim the content after the first dot? So gke-wi-collector-test-default-pool-5a7b28f7-nxwt.c.my-project.internal would become gke-wi-collector-test-default-pool-5a7b28f7-nxwt?

[dashpole]

We could try that, since the default hostname is the instance name. But it is possible for users to override the hostname of their VMs, so it isn't guaranteed to work correctly. Given that node.name is available through the downward API, is guaranteed to be correct, and we have other things we need to get through the downward API, I'm inclined to start with that approach.