Function spmv_axpby 

pub fn spmv_axpby(
    matrix: &CsrMatrix,
    x: &[f64],
    y: &mut [f64],
    alpha: f64,
    beta: f64,
) -> Result<()>

SpMV with y = alpha * A * x + beta * y (BLAS-style).

This allows accumulating results without allocating new vectors.