fix: number of rounds for mimc

[yelhousni]

No description provided.

[yelhousni]

For bls12-377, I changed it to:

• permutation exponent e=17 (instead of -1)
• number of rounds r=62 (log(r)/log(17)

This follows Penumbra reasoning (doc, audit) for the Poseidon hash. Out-circuit x^17 can be done in 4 squaring and 1 multiplication (much faster than an inverse) and in-circuit in 5 R1CS constraints per round (62*5=310), which is faster than BN254 (110*3=330) and BLS12-381 (111*3=333).

Corresponding PR on gnark side: Consensys/gnark#453