Sadly, the streaming function `extractRawInterestBits` is not faster than the non-streaming version. Worthwhile investigation if it is possible to get better performance using SIMD in straight C.