fix: close leaked replica in TestShip_DegradedDoesNotSilently

The test used createSyncAllPair(t) but discarded the replica
return value, leaving the volume file open. On Windows this
caused TempDir cleanup failure. All 7 CP13-1 baseline FAILs
now PASS.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

weed/storage/blockvol/sync_all_protocol_test.go

@@ -462,8 +462,9 @@ func TestWalRetention_RequiredReplicaBlocksReclaim(t *testing.T) {
 // This is acceptable for best_effort but problematic for sync_all because
 // the primary loses track of the replica gap size.
 func TestShip_DegradedDoesNotSilentlyCountAsHealthy(t *testing.T) {
-	primary, _ := createSyncAllPair(t)
+	primary, replica := createSyncAllPair(t)
 	defer primary.Close()
+	defer replica.Close()
 
 	// Point shipper at dead address — will degrade on first Ship.
 	primary.SetReplicaAddr("127.0.0.1:1", "127.0.0.1:2")