Function mla_decode_token 

pub fn mla_decode_token(
    q: &mut [f32],
    k: &mut [f32],
    v: &[f32],
    g: &[f32],
    kv: &mut MlaKv,
    cfg: &MlaConfig,
    pos: usize,
    theta: f32,
) -> Vec<f32>

One decode token: optional RoPE, append K/V, SDPA, optional output gate. Projections stay in mla_mixer. CUDA k3_mla_* matches this order.