amazon: add a Ready() signal for ProcessEvents' LISTEN registration

Tests (and any other caller) previously had no way to know when the
Postgres LISTEN behind the reactive event-processing loop had actually
registered, forcing a guessed sleep before relying on it. Mocks now
exposes Ready() <-chan struct{}, closed once listenForNotifications
successfully issues LISTEN. Purely additive - ProcessEvents' signature
is unchanged.

Updates the integration tests to wait on Ready() instead of a flat
sleep, which also cut TestProcessEvents_ReactsToNotification's runtime
from ~0.25s to ~0.06s with no flakes across repeated runs.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XEDaCB7C2NEBgyvqEtZuxY

domains/amazon/mock_test.go

@@ -184,14 +184,11 @@ func TestProcessEvents_ReactsToNotification(t *testing.T) {
 		errCh <- m.ProcessEvents(ctx)
 	}()
 
-	// ProcessEvents registers its Postgres LISTEN synchronously before
-	// entering its main loop, but that registration still races against
-	// this goroutine actually getting scheduled - the production code
-	// exposes no "listening now" signal a caller (or a test) can wait
-	// on. A short sleep is the only lever available from outside; see
-	// the accompanying report for why that's a real testability/design
-	// gap, not just a test-flakiness workaround.
-	time.Sleep(200 * time.Millisecond)
+	select {
+	case <-m.Ready():
+	case <-time.After(5 * time.Second):
+		t.Fatal("ProcessEvents() did not become ready (LISTEN registered) within 5s")
+	}
 
 	insertRawAmazonEvent(t, pool, shopID, "evt-1")
 
@@ -234,7 +231,11 @@ func TestProcessEvents_ShutsDownOnContextCancel(t *testing.T) {
 		errCh <- m.ProcessEvents(ctx)
 	}()
 
-	time.Sleep(50 * time.Millisecond) // let it reach its first select
+	select {
+	case <-m.Ready():
+	case <-time.After(5 * time.Second):
+		t.Fatal("ProcessEvents() did not become ready (LISTEN registered) within 5s")
+	}
 	cancel()
 
 	select {