v0.4.0

0.4.0 — 2026-04-13

Audit-log hardening. The audit.Logger has existed since v0.1.0 but was
never actually instantiated by cmd/main.go — a gap this release closes,
alongside a retention policy and metrics so the audit_log table
doesn't grow forever.

Added

• cmd/main.go flags: --audit-dsn (or TSDB_AUDIT_DSN env),
  --audit-retention-days. When DSN is set, the logger is opened on
  startup and handed to the REST API server, so
  GET /api/clusters/:name/audit and the auto-recorded entries on
  create/delete/backup routes actually work.
• audit.Logger.Prune(ctx, before) deletes rows older than the given
  cutoff and returns the count.
• audit.Logger.Start(ctx) is a manager.Runnable pruner. Runs every
  hour (configurable via PruneInterval) plus once on startup so the
  effect is visible immediately. No-op when RetentionDays == 0.
• Metrics: tsdb_operator_audit_record_total{cluster,result},
  tsdb_operator_audit_prune_total, tsdb_operator_audit_rows.
• Helm chart: features.audit, audit.dsnSecretRef,
  audit.retentionDays. DSN is plumbed via TSDB_AUDIT_DSN env from
  a Secret.
• Tests for Prune, RowCount, and pruner-start no-op using
  DATA-DOG/go-sqlmock.