feat: add read-only system metadata collector (experimental)

[rokernel]

Summary

• add a new system_collector that exports:
  • sonic_system_identity_info
  • sonic_system_software_info
  • sonic_system_uptime_seconds
  • collector health metrics (collector_success, scrape_duration_seconds, cache_age_seconds)
• implement source fallback order to support mixed SONiC versions:
  1. Redis (DEVICE_METADATA|localhost, CHASSIS_INFO|chassis 1)
  2. Read-only files (/etc/sonic/sonic_version.yml, /host/machine.conf, /etc/hostname, /proc/uptime)
  3. Optional read-only command fallbacks (show platform summary --json, show version, show platform syseeprom)
• enforce safety and stability controls:
  • read-only design (no Redis/file writes)
  • command allowlist
  • command timeout + max output bytes
  • cached refresh path so /metrics stays responsive
  • missing values mapped to unknown
• add debug-level validation logs to explain:
  • missing expected fields
  • which source populated a field
  • fallback sources ignored due to priority

Stability Decision

system_collector is still experimental and is now disabled by default for stability.

• SYSTEM_ENABLED default: false
• Explicit opt-in required:
  • SYSTEM_ENABLED=true ./sonic-exporter

Documentation Updates

README now clearly documents:

• experimental status and default-off behavior
• enable command
• metrics exposed by system collector
• source order and fallback strategy
• read-only/security behavior
• debug logging behavior
• all SYSTEM_* configuration flags

Data Hygiene

• test fixtures and examples are anonymized (hostname/serial/model/build metadata)
• no real confidential values added to repository

Verification

Local

• go test ./...
• go build ./...
• go run github.com/golangci/golangci-lint/cmd/golangci-lint@v1.64.8 run

Runtime (both SONiC versions)

Validated on:

• exoadmin@192.168.240.124 (SONiC 202012)
• exoadmin@192.168.240.141 (SONiC OS12)

Checks performed:

1. default behavior (SYSTEM_ENABLED unset): no sonic_system_* metrics exposed
2. enabled behavior (SYSTEM_ENABLED=true):
   • sonic_system_identity_info present
   • sonic_system_software_info present
   • sonic_system_uptime_seconds present
   • sonic_system_collector_success 1
3. debug behavior: clear logs for expected missing fields and fallback decisions