KubeVirt provider should honor the explicitly selected instancetype kind

The KubeVirt provider's `defaultVMInstanceType` function currently ignores the `Kind` supplied in `InstancetypeMatcher` and discovers the resource by name.

The lookup order is currently: Namespaced `VirtualMachineInstancetype` and then Cluster scoped `VirtualMachineClusterInstancetype`. This happens even when the provider specification explicitly contains:

```yaml
instancetype:
name: h200-worker
kind: VirtualMachineClusterInstancetype
```

If both a namespaced VirtualMachineInstancetype/h200-worker and a cluster VirtualMachineClusterInstancetype/h200-worker exist, machine-controller selects the namespaced object despite the explicitly submitted cluster scoped kind