quotient ring exponentiation

[odraccir1]

Add quotient-ring arithmetic needed by polynomial factorization:

• Add mulmod and squaremod in a dedicated modular-arithmetic header.
• Add binary powmod with large-integer exponent support.
• Support reusable precomputed divisor contexts and output aliasing.
• Add generic prime- and extension-field order calculation.

Compared with FLINT, powmod is approximately 6–9× slower, while large squaremod reaches roughly 12×. This is wider than the previously measured 3–5× polynomial-multiplication gap. The gap widens because every modular multiplication performs a full product plus two nominal low products during preinverse reduction. The current mixed-radix backend implements those low products using complete FFT convolutions and then discards unwanted coefficients. FLINT instead uses specialized truncated and middle-product algorithms.

So again the most obvious path to closing the gap is to implement

• A performant large-integer polynomial backend.
• Kronecker packing/unpacking for Fq/Fq12.
• Specialized full, truncated, and middle products.

Related to #114