Add Metrics

[andrew-d]

The service should expose metrics (likely Prometheus) on the debug port so that it's clearer if it's working as expected. Ideas:

• Gauge with the last time we successfully fetched the Proxmox inventory
• Gauge with number of discovered nodes/LXCs/VMs
• Number of DNS requests, split by TCP/UDP and A/AAAA
• Inventory fetch latency
• Success/failure counters for fetching things from Proxmox (e.g. an inventory_fetch_failure metric with a reason attribute)