feat(comm): T1B.3 — central-side gRPC site-command transport seam (default Akka)

Extract the central→site send path in CentralCommunicationActor behind a new
ISiteCommandTransport, selected by ScadaBridge:Communication:SiteTransport
(Akka | Grpc, default Akka — rollback = flip the flag). CommunicationService's
27 commands, SiteCallAuditActor's 2 parked relays and DebugStreamBridgeActor's
subscribe/unsubscribe are untouched; the seam sits below SiteEnvelope.

- AkkaSiteTransport: today's per-site ClusterClient path extracted verbatim
  (the _siteClients lookup + ClusterClient.Send with the reply-to sender
  preserved, and the "no client ⇒ warn + drop, caller's Ask times out" path).
- GrpcSiteTransport: dials the site SiteCommandService (T1B.1 proto client) via
  SiteCommandDtoMapper, PSK + x-scadabridge-site on the channel through
  ControlPlaneCredentials, per-command deadlines set EQUAL to today's
  CommunicationService Ask timeouts (per-command, not per-group: DeploymentState
  query and TriggerSiteFailover use QueryTimeout; the two parked relays map to
  QueryTimeout so SiteCallAudit's inner RelayTimeout 10s < 30s ordering holds;
  WaitForAttribute keeps its dynamic Timeout + IntegrationTimeout).
- SitePairChannelProvider: per-site A/B channel pair with sticky failover
  (flip only on Unavailable — NEVER on DeadlineExceeded, a write/deploy/failover
  may have run), background failback probe to the preferred node with 1s→60s
  doubling backoff, PSK invalidation on site removal. Fed by the SAME DB refresh
  loop (extended to carry GrpcNodeA/GrpcNodeBAddress) — no second poll.

Tests: actor-with-substitute-transport (routing, Ask-reply plumbing, per-site
lifecycle across refreshes), ResolveDeadline pinned to each command's current
Ask timeout, and GrpcSiteTransport/SitePairChannelProvider over dual in-process
TestServers (PSK+header+deadline attached, Unavailable failover + stickiness,
failback to preferred, no-retry-on-DeadlineExceeded). Proto csproj untouched
(no active <Protobuf> item). Full solution builds 0 warnings; Communication.Tests
607 green with Akka default.

src/ZB.MOM.WW.ScadaBridge.Communication/Actors/ISiteCommandTransport.cs

@@ -0,0 +1,43 @@
+using Akka.Actor;
+
+namespace ZB.MOM.WW.ScadaBridge.Communication.Actors;
+
+/// <summary>
+/// The central→site command-send seam, injected into <see cref="CentralCommunicationActor"/>
+/// below the <see cref="SiteEnvelope"/> handler. Exactly one implementation is active per node,
+/// chosen by <c>ScadaBridge:Communication:SiteTransport</c>:
+/// <list type="bullet">
+///   <item><see cref="AkkaSiteTransport"/> — today's per-site <c>ClusterClient</c> path (default).</item>
+///   <item><see cref="ZB.MOM.WW.ScadaBridge.Communication.Grpc.GrpcSiteTransport"/> — the site
+///     <c>SiteCommandService</c> gRPC plane.</item>
+/// </list>
+/// The producers above the seam (<c>CommunicationService</c>'s 27 commands, <c>SiteCallAuditActor</c>'s
+/// 2 parked relays, and <c>DebugStreamBridgeActor</c>'s subscribe/unsubscribe) are unchanged — they
+/// still <c>Ask</c>/<c>Tell</c> a <see cref="SiteEnvelope"/> to the actor, which delegates here.
+/// </summary>
+/// <remarks>
+/// Both members run on the actor thread (from the <see cref="SiteEnvelope"/> and
+/// <c>SiteAddressCacheLoaded</c> handlers), so implementations need no internal synchronisation for
+/// their own per-site bookkeeping beyond what a background failback loop requires.
+/// </remarks>
+public interface ISiteCommandTransport
+{
+    /// <summary>
+    /// Routes <paramref name="envelope"/>'s message to its site. Any reply the site produces is
+    /// delivered to <paramref name="replyTo"/> — for an <c>Ask</c> that is the temporary ask actor
+    /// (completing the caller's task); for a <c>Tell</c>-with-sender (the debug bridge) that is the
+    /// originating actor. A message with no route (an unknown site) is warned and dropped so the
+    /// caller's <c>Ask</c> times out, exactly as today's ClusterClient path behaves.
+    /// </summary>
+    /// <param name="envelope">The site-addressed command envelope.</param>
+    /// <param name="replyTo">Where a reply (or a <see cref="Status.Failure"/>) is delivered.</param>
+    void Send(SiteEnvelope envelope, IActorRef replyTo);
+
+    /// <summary>
+    /// Reconciles per-site transport resources (ClusterClients for Akka, channel pairs for gRPC)
+    /// against the freshly loaded site set. Called once per DB refresh tick with the same cache
+    /// message the actor already receives — the ONE DB-poll loop feeds both transports.
+    /// </summary>
+    /// <param name="cache">The loaded site address cache (Akka contacts + gRPC endpoints + known ids).</param>
+    void ReconcileSites(SiteAddressCacheLoaded cache);
+}