Prometheus /metrics endpoint

[proofmancer]

Why

Once `faucet.ligate.io` is live, ops needs to monitor: drip success rate, drip latency, rate-limiter rejections, signer-key signing latency, sequencer submission failures. Today the faucet logs to stdout but exposes no scrapable metrics surface.

What

Add a Prometheus `/metrics` endpoint mirroring the chain repo's pattern (`crates/rollup/src/metrics.rs`):

• `faucet_drips_total{outcome="success|rate_limited|invalid_address|signer_error|submit_error"}` counter
• `faucet_drip_duration_seconds` histogram (request → tx_hash returned)
• `faucet_signer_nonce` gauge (current atomic nonce)
• `faucet_addresses_dripped` gauge (mirror of `/faucet/status` but in Prometheus shape)
• `faucet_rate_limit_evictions_total` counter

Bind on a separate port (e.g. `:9101`) so the public `:8080` doesn't expose internals.

Why pre-devnet

Without this, when someone reports "the faucet isn't dripping" we have no signal except logs. With this, we can put a Grafana panel on `grafana.ligate.io` and see the failure shape immediately.

Done when

• `/metrics` endpoint serves the listed metrics
• README documents the bind address + scrape config
• Wired into the existing `docker-compose.yml` deploy