Function paged_decode_attn_bf16k_turbo4v 

pub fn paged_decode_attn_bf16k_turbo4v(
    gpu: &dyn GpuBackend,
    kernel: KernelHandle,
    q: DevicePtr,
    k_cache: DevicePtr,
    v_cache: DevicePtr,
    output: DevicePtr,
    block_tables: DevicePtr,
    seq_lens: DevicePtr,
    max_blocks_per_seq: u32,
    num_seqs: u32,
    num_q_heads: u32,
    num_kv_heads: u32,
    head_dim: u32,
    block_size: u32,
    inv_sqrt_d: f32,
    q_stride: u32,
    v_block_stride_bytes: u64,
    v_data_section_bytes: u64,
    sliding_window: u32,
    stream: u64,
) -> Result<()>

Paged decode attention for Bf16K + Turbo4V asymmetric KV cache.

K read as BF16 NHD, V as 4-bit Lloyd-Max packed bytes + FP8 per-group scale (sparse V on batched + remainder paths).