[telemetry] Prometheus metrics endpoint (liquidations, profit, scan latency, chain health)

[obchain]

Context

Visibility is non-negotiable for a live bot. Prometheus + Grafana is self-hosted, zero marginal cost, and integrates with every standard alerting tool. This issue covers the exposition side; the Grafana dashboards are part of the deploy issue.

Scope

• Create crates/charon-telemetry/ crate
• /metrics HTTP endpoint on a configurable port (default 9090)
• Counters: charon_liquidations_total, charon_liquidation_failures_total, charon_scans_total
• Gauges: charon_positions_tracked, charon_latest_block, charon_chain_ws_connected
• Histograms: charon_scan_duration_seconds, charon_build_duration_seconds
• Labels: chain, protocol, bucket (healthy / near-liq / liquidatable)
• prometheus crate (or metrics + metrics-exporter-prometheus)

Acceptance criteria

• curl localhost:9090/metrics returns Prometheus-formatted text
• All metrics emit during a normal listen run
• Counter increments are observable across multiple scans
• Histograms capture realistic latency distributions
• No metric-cardinality explosions (label values bounded)

References

• PRD section 7a (Telemetry & Alerting)

[obchain]

Superseded — Prometheus exporter shipped in feat/22-prometheus-metrics.