[Nvidia GPU] GPU-Specific Overview hostname control uses gpu.labels.Hostname

### Integration Name

NVIDIA GPU [packages/nvidia_gpu]

### Integration Version

0.4.1

### What did you do?

Install NVIDIA GPU Monitoring 0.4.1 and open the shipped dashboard `[Metrics Nvidia GPU] GPU-Specific Overview`. Use the top-level **GPU Hostname** control to filter by host.

### What did you see?

The control is bound to `gpu.labels.Hostname` and returns no values. Indexed documents use `gpu.labels.hostname` (lowercase `h`). The leftover is still on `main`:

https://github.com/elastic/integrations/blob/fce6bf9b63d6f086e2faa877c0f598bdc7eed8e0/packages/nvidia_gpu/kibana/dashboard/nvidia_gpu-bac121fe-6cd9-48a7-a349-f52ffa42d56b.json#L21

```json
"fieldName": "gpu.labels.Hostname",
"title": "GPU Hostname",
```

This was introduced by the field rename in https://github.com/elastic/integrations/pull/14081 (package 0.2.0).

### Workaround

Clone the dashboard and point the control at `gpu.labels.hostname`.

Open packages/nvidia_gpu/kibana/dashboard/nvidia_gpu-bac121fe-6cd9-48a7-a349-f52ffa42d56b.json around line 21 and inspect the GPU Hostname control. Compare its field name with the indexed field described in the issue, then verify the shipped dashboard control returns host values when filtering.

Written by the indexing model from the issue text.