Move ChunkId hashing from producer into rayon pool for parallel computation

The producer thread was bottlenecked on sequential BLAKE2b-256 hashing
(~1 GB/s/core), limiting throughput while rayon workers sat mostly idle.
Now all per-byte CPU work (hash + compress + encrypt) runs in the rayon
pool, letting the producer focus on I/O + FastCDC only.

Author: m3nu