Function linear_bf16 

pub fn linear_bf16(
    x: &[f32],
    w: &[f32],
    rows: usize,
    in_dim: usize,
    out_dim: usize,
) -> Vec<f32>

2026-09-25: Linear on bf16 operands: y[i][o] = bf16(sum_d x[i][d] * w[o][d]), f32 accumulate. w is [out, in] row-major.