The newish SBCL parallel gc is fantastic and uses no additional memory during gc