[None][feat] Add triton paged attention for AutoDeploy (NVIDIA#12642)

Signed-off-by: Chenghao Zhang <211069071+nvchenghaoz@users.noreply.github.com>
Co-authored-by: Suyog Gupta <41447211+suyoggupta@users.noreply.github.com>

Author: nvchenghaoz

tensorrt_llm/_torch/auto_deploy/custom_ops/attention/__init__.py

@@ -22,7 +22,7 @@
 - trtllm_attention: TRT-LLM thop.attention-based optimized attention
 - triton_attention: Triton-based attention implementations
 - triton_attention_with_kv_cache: Triton attention with KV cache support
-- triton_attention_with_paged_kv_cache: Triton attention with paged KV cache
+- triton_paged_attention: Triton paged attention (two-stage flash-decode) with HND layout
 - onnx_attention: Placeholder ops for ONNX export of attention mechanisms
 """
 
@@ -34,5 +34,6 @@
     "triton_attention",
     "triton_attention_with_kv_cache",
     "triton_attention_with_paged_kv_cache",
+    "triton_paged_attention",
     "onnx_attention",
 ]