Function masked_mat_vec_mul 

pub fn masked_mat_vec_mul(
    a_hat: &[[[i32; 256]; 7]; 8],
    y_hat_m: &[MaskedPoly],
    k: usize,
    l: usize,
    out: &mut [MaskedPoly],
)

Masked matrix-vector multiplication in the NTT domain: for each output row i, compute sum_j (A_hat[i][j] · y_hat_m[j]) as a masked accumulator.

Since A_hat is public (derived from the public seed rho) and the NTT is linear, each share is multiplied independently by the same public matrix and accumulated into the corresponding output share — no secret×secret operation occurs, so first-order shares remain sufficient.

y_hat_m must have length ≥ l (the first l masked polynomials are consumed). out must have length ≥ k.