Function mla_cache_assemble 

pub fn mla_cache_assemble(
    gpu: &dyn GpuBackend,
    kernel: KernelHandle,
    kv_latent: DevicePtr,
    k_rope: DevicePtr,
    k_cache: DevicePtr,
    v_cache: DevicePtr,
    kv_lora: u32,
    rope: u32,
    mla_cache_dim: u32,
    stream: u64,
) -> Result<()>

MLA cache assembly: fuse [kv_latent|k_rope]→K and [kv_latent|zeros]→V into 1 kernel.