Try parallelizing hashing over multiple files

[casey]

This is apropos of the discussion in BLAKE3-team/BLAKE3#205.

We should try parallelizing single-threaded reads over multiple files. Single-threaded reads which use update_reader would allow for more granular progress bars, and might wind up being more efficient overall.

[rot13maxi]

I like this idea. Spin up $CORE_COUNT number of threads, and show that many progress bars. Would be cool to see 8 or 16 progress bars whipping by. Since there's already a transitive dependency on Rayon, could experiment with their thread pool rather than rolling something from scratch (using Thread::spawn and mpsc channels)