feat(observability): CountDIDs + DID/pending/peer counts in status + /stats (#103)

[LiranCohen]

Summary

Closes audit finding #14, and adds the directly-requested capability: how many DIDs has this node indexed? The node exposed no counts — no DID total, the status CLI omitted pending and peer counts, and there was no machine-readable stats endpoint. The two most common silent degradations (IPFS down → growing pending backlog; zero peers → sync stall) had no easy signal.

What's added

• store.CountDIDs() (int, error) — distinct projected DID suffixes (sqlite COUNT(DISTINCT suffix); bbolt counts distinct prefixes in the sorted did_fwd bucket). Reorg-coupled.
• p2p.Client.PeerCount() int.
• ion-node status now prints dids and pendingRetryable.
• The live start status log adds peers, dids, pendingRetryable.
• GET /stats (agent-native parity with the CLI) → {dids, pendingRetryable, committedHeight, processedHeight, fullyResolved}. Registered only when a StatsProvider is configured (serve wires the store via Handler.WithStats).

How to use

ion-node status                          # network=… committedHeight=… indexedAnchors=… dids=… pendingRetryable=… datadir=…
curl -s http://127.0.0.1:8080/stats      # {"dids":…,"pendingRetryable":…,"fullyResolved":…,…}

Caveat

dids reflects only locally-resolved content, so it can undercount while pendingRetryable > 0 (anchored content not yet fetched from IPFS). fullyResolved/pendingRetryable signal completeness.

Tests

• TestCountDIDs — distinct suffixes (not rows: a DID with several ops counts once), reorg-coupled, both engines.
• TestStatsEndpoint — /stats reports counts; returns 404 without a provider.

go test -race ./... green (26 packages).

Post-Deploy Monitoring & Validation

• What to watch: peers (should track --outbound), pendingRetryable (should trend to 0 when IPFS is healthy), and dids (monotonic up as sync progresses). Scrape GET /stats from monitoring.
• Failure signal / trigger: peers near 0 (sync stall), or pendingRetryable growing without bound (IPFS unreachable).
• Window/owner: ongoing operational dashboards.

Closes #103

🤖 Generated with Claude Code