feat(metrics): Prometheus-compatible /metrics endpoint

[iceice400]

Summary

• Adds optional GET /metrics in Prometheus text 0.0.4 format (zero-dependency writer)
• Exposes packet counts, node totals, RSSI/SNR averages, noise floor, relay stats, per-channel duty (ToA), SX1302 CRC counters, and uptime from in-memory state
• Disabled by default (metrics.enabled: false); optional metrics.require_auth for LAN scrapers

Why

Operators running Prometheus/Grafana on the LAN need a standard scrape target without polling multiple JSON API routes. Default-off keeps upgrade behaviour unchanged.

Type

• Feature
• Bug fix
• Docs
• Refactor
• UI
• Installer
• Region support
• Hardware change

Testing

• python -m unittest tests.test_metrics_routes -v (6 tests, pass)
• Tested on hardware with Prometheus scrape
• Dashboard tested

Test plan

1. Fresh install / upgrade with default config → GET /metrics returns 404
2. Set metrics.enabled: true, require_auth: false → scrape returns text/plain with meshpoint_packets_session_total, meshpoint_noise_floor_dbm, etc.
3. Confirm no PSK/token values appear in metric labels or sample values
4. With require_auth: true, unauthenticated scrape returns 401
5. Normal dashboard and packet pipeline unchanged

Relay / TX impact

None. Read-only exposition endpoint; no pipeline or relay changes.

Config

metrics:
  enabled: false
  require_auth: true

Documented in docs/CONFIGURATION.md with example prometheus.yml scrape job.

AI-assisted?

Implemented with AI assistance; unit tests run locally.