GliDe with a CaPE: A Low-Hassle Method to Accelerate Speculative Decoding