IP4 InstancesV2 address should not be hardcoded to ETH0_IP

[sk4zuzu]

The problem seems to be https://github.com/OpenNebula/cloud-provider-opennebula/blob/master/pkg/cloud/instancesv2.go#L94

In case of multiple VNETs are attached to the node/VM users should have ability to pick which IP(s) are put inside k8s in the node metadata or/and those IPs should be dynamically inferred from attached VNETs/cloud-config.

[balazsbme]

Some more information about the issue:

Currently the cloud provider only handles one IP per VM reliably. See this error in pod logs:

"Unhandled Error" err="error syncing 'rke2-capone-workload-cell-site-md01-6qmdj-c8d7f': failed to get node modifiers from cloud provider: provided node ip for node \"rke2-capone-workload-cell-site-md01-6qmdj-c8d7f\" is not valid: failed to get node address from cloud provider that matches ip: 10.0.1.155, requeuing" logger="UnhandledError"

As a concequence, we cannot create VMs with CAPONE that have more than 1 NIC attached. Note that with PCI passthrough of a NIC (e.g. VF for SRIOV) is not handled by code part, so it wont be affected, its netwroking is not configured with Opennebula context scripts.