Improve the blake2s hash circuit

[XuyangSong]

Now, we have a straightforward blake2s hash circuit implementation, which could be more efficient. A lookup table with 8 bits would help.

• convert the range check to lookup
• convert the xor, shift, and other operations to lookup, inspired by halo2 sha256 implementation.
• other potential tricks to improve it, from Risc0 and Miden, perhaps