feat(metrics): enrich device metrics with pod mapping and health status

[rebel-mskim]

Motivation

Rebellions’ NPU exporter currently emits only bare device metrics. Operators need better visibility into which Kubernetes pod owns each device and whether the device is healthy, including driver/firmware metadata, so they can troubleshoot workloads quickly.

Summary of Changes

• add a kubelet-based pod resource mapper, node-name detection, and shared label definitions so all collectors can tag metrics with pod/namespace/container/hostname and device version info
• extend the daemon client and existing collectors to populate the enriched labels, and introduce a device health collector plus factory wiring to export per-device health status gauges

Technical Details

• wire the new PodResourceMapper and node name through config → collector factory → hardware/memory/utilization collectors, ensuring Gauges use the expanded label set consistently
• augment the daemon client with version info and GetTotalInfo lookups so DeviceStatus carries driver/firmware/SMC versions and health status, which the new DeviceHealthCollector exposes via RBLN_DEVICE_STATUS:HEALTH