[Feature]: Add dedicated metrics for Snapshot feature

[zhuwenxing]

Is there an existing issue for this?

• I have searched the existing issues

Is your feature request related to a problem? Please describe.

The Snapshot feature (introduced in #44358) currently only uses generic Proxy-level metrics (milvus_proxy_function_call_total and milvus_proxy_req_latency). These metrics only track API call counts and latency, but lack snapshot-specific operational insights.

Current limitations:

1. Cannot monitor the total number of snapshots across collections
2. Cannot track snapshot storage consumption
3. Cannot observe restore operation progress in real-time via metrics
4. Cannot measure snapshot creation/restore duration at the DataCoord level
5. No visibility into snapshot-referenced data that cannot be garbage collected

This makes it difficult to:

• Set up alerting for snapshot storage growth
• Monitor restore job progress via Grafana dashboards
• Understand the storage impact of snapshots on object storage costs
• Troubleshoot snapshot-related performance issues

Describe the solution you'd like

Add dedicated Prometheus metrics for the Snapshot feature:

Snapshot Inventory Metrics

Metric Name	Type	Labels	Description
milvus_snapshot_total	Gauge	collection_id, db_name	Total number of snapshots
milvus_snapshot_storage_bytes	Gauge	collection_id, snapshot_name	Storage size of snapshot data
milvus_snapshot_referenced_storage_bytes	Gauge	collection_id	Storage size of data referenced by snapshots (cannot be GC'd)

Snapshot Operation Metrics

Metric Name	Type	Labels	Description
milvus_snapshot_create_duration_seconds	Histogram	collection_id, status	Time to create a snapshot
milvus_snapshot_restore_duration_seconds	Histogram	collection_id, status	Time to restore a snapshot
milvus_snapshot_restore_progress_ratio	Gauge	job_id, snapshot_name	Restore progress (0.0 - 1.0)
milvus_snapshot_restore_jobs_total	Gauge	state	Number of restore jobs by state (pending/in_progress/completed/failed)

Snapshot Error Metrics

Metric Name	Type	Labels	Description
milvus_snapshot_operation_errors_total	Counter	operation, error_type	Total snapshot operation errors

Describe an alternate solution

1. Extend existing metrics: Add operation_type=snapshot_* labels to existing DataCoord metrics instead of creating new metric families.

2. Expose via API only: Keep metrics lightweight and expose detailed snapshot statistics only via DescribeSnapshot API responses, letting users build custom exporters.

Anything else? (Additional Context)

• Related Feature Issue: [Feature]: Add Snapshot Functionality for Collections #44358
• Current metrics location: pkg/metrics/proxy_metrics.go
• Snapshot implementation: internal/proxy/snapshot_impl.go, internal/datacoord/snapshot*.go

User Guide Note: The snapshot user guide mentions "Monitoring: Track snapshot creation times and storage usage" as a best practice, but currently there's no built-in way to do this via metrics.