feat(audit-log): surface partition-purge failure as health event + counter

src/ZB.MOM.WW.ScadaBridge.AuditLog/Central/AuditLogPurgeFailedEvent.cs

@@ -0,0 +1,29 @@
+namespace ZB.MOM.WW.ScadaBridge.AuditLog.Central;
+
+/// <summary>
+/// Symmetric failure twin of <see cref="AuditLogPurgedEvent"/>. Published on the
+/// actor-system EventStream by <see cref="AuditLogPurgeActor"/> whenever a purge
+/// step throws — a partition switch-out, a per-channel override DELETE, or the
+/// boundary enumeration that precedes them. A silently failing retention job is a
+/// health-surface condition, not just a log line, so this event lets the central
+/// health collector + ops dashboards observe purge failures without coupling to
+/// the actor (arch-review 04, S2).
+/// </summary>
+/// <param name="MonthBoundary">
+/// The pf_AuditLog_Month lower-bound boundary whose switch-out failed. When the
+/// failure occurs BEFORE any specific boundary is selected — i.e. the pre-purge
+/// boundary enumeration threw — this is <see cref="System.DateTime.MinValue"/> as a
+/// sentinel for "enumeration phase, no boundary in hand".
+/// </param>
+/// <param name="Error">
+/// The failure's message (<see cref="System.Exception.Message"/>). The full
+/// exception is logged at Error on the actor's logger; this carries the short form
+/// for surfacing on health/ops without leaking a full stack trace onto the stream.
+/// </param>
+/// <param name="ElapsedMilliseconds">
+/// Wall-clock time spent on the failed step before it threw, in milliseconds.
+/// </param>
+public sealed record AuditLogPurgeFailedEvent(
+    DateTime MonthBoundary,
+    string Error,
+    long ElapsedMilliseconds);