> In this implementation, to maximise performance, we consider each individual byte of the file separately, and look it up in π.

Considering each individual bit separately would be even more performant: you only need the indexes 2 and 33, and there is an efficient mapping of those to the bits in storage.