Add resource capacity monitoring metrics

[sjmiller609]

Summary

• add a configurable metrics refresh interval for cached resource capacity monitoring
• add cached OTel resource capacity gauges for core resources, disk/image storage, and GPU slots
• start the resource monitoring refresh loop during API startup while keeping /resources live

Testing

• go test ./cmd/api/config ./lib/resources

Notes

• A broader go test ./cmd/api/... run is currently blocked in this worktree by missing embedded binaries (vz-shim/vz-shim and guest_agent/guest-agent).

---

Note

Medium Risk
Adds a new background monitoring loop at API startup that periodically snapshots host resource status for OTel metrics; incorrect intervals or snapshot failures could impact observability and introduce extra load, but core request paths are mostly unchanged.

Overview
Adds cached OpenTelemetry resource-capacity metrics driven by a new ResourceManager.StartMonitoring refresh loop that periodically snapshots GetFullStatus and publishes observable gauges for capacity/limits/allocations, disk/image storage breakdown, and GPU slot availability.

Introduces a new metrics.resource_refresh_interval config (default 120s) with validation and examples, wires the refresh loop into API startup, and updates/extends tests (new lib/resources/monitoring_test.go, plus small stability tweaks in integration/registry tests around async waits and timeouts).

^{Written by Cursor Bugbot for commit 45bf9f4. This will update automatically on new commits. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

[hiroTamada]

solid design — snapshot-based caching with a single OTel callback is clean. one minor comment on goroutine durability.