fix(cache): harden cache backends, resilience, recovery, and transports

- RESP -ERR no longer closes healthy conn / aborts batch; global cap replacement-aware
- global health-check pass serialized (latent data race); MemoryMonitor via ResourceManager
- token-exchange retry, open-circuit rejections counted as failures
- TLS-posture matching transport reuse (no silent InsecureSkipVerify handoff)

internal/cache/review_r149_test.go

@@ -0,0 +1,30 @@
+package cache
+
+import (
+	"testing"
+	"time"
+)
+
+// TestCacheClose_NilStopCleanupNoPanic guards the R149 fix in Close():
+// stopCleanup is only allocated when EnableAutoCleanup && CleanupInterval > 0,
+// but Close() previously closed it whenever EnableAutoCleanup was set. A
+// cache built with EnableAutoCleanup=true and CleanupInterval left at its
+// zero value therefore called close(nil) and panicked during shutdown.
+func TestCacheClose_NilStopCleanupNoPanic(t *testing.T) {
+	c := New(Config{EnableAutoCleanup: true}) // CleanupInterval defaults to 0
+	if err := c.Close(); err != nil {
+		t.Fatalf("Close() with CleanupInterval=0 must not panic and return nil, got: %v", err)
+	}
+}
+
+// TestCacheClose_WithCleanupIntervalClosesOnce ensures the normal path
+// (cleanup started) still closes cleanly exactly once.
+func TestCacheClose_WithCleanupIntervalClosesOnce(t *testing.T) {
+	c := New(Config{EnableAutoCleanup: true, CleanupInterval: time.Minute})
+	if err := c.Close(); err != nil {
+		t.Fatalf("Close() with a started cleanup goroutine must succeed, got: %v", err)
+	}
+	if err := c.Close(); err == nil {
+		t.Fatal("second Close() must report already-closed")
+	}
+}