Faster squaring in Fp

[Nashtare]

This paper introduces a faster way than Tonelli-Shanks to compute the square root of a field element if the field characteristic is highly 2-adic.
According to the table 1 page 12, the number of field operations in the case of Fp (with 2-adicity 32) would go from 311 to 134.5 when switching algorithm.

[Nashtare]

Note: This will not impact point operations requiring to compute a square root in Fp6, as Fp6::sqrt() internally relies on Fp3::sqrt() which directly implements Tonelli-Shanks.