pub fn linear_tensor_vars(rows: usize) -> usize
Challenge count for outer_r_lin. The tensor batches the affine rows at k / |F|, not the 1 / |F| of one challenge per row.
outer_r_lin
k / |F|
1 / |F|