[Performance]Optimize DeepSeekOCR2 RelPosAttention and CustomQwen2Decoder (#7737)

### What this PR does / why we need it?
Optimize DeepSeekOCR2 RelPosAttention and CustomQwen2Decoder and add doc
for DeepSeekOCR2.md

### Does this PR introduce _any_ user-facing change?

### How was this patch tested?
- vllm 0.18.0
- vllm-ascend main

1. _create_custom_4d_mask during 141ms49us620ns -->
_create_npu_optimized_mask during 1ms227us780ns
2. convd2d : 27ms --> matmul <1ms
3. relposattention：sdpa->prompt_flash_attention

---------

Signed-off-by: Wangbei25 <wangbei41@huawie.com>
Signed-off-by: Wangbei25 <wangbei41@huawei.com>
Co-authored-by: Wangbei25 <wangbei41@huawie.com>

vllm_ascend/utils.py

@@ -604,7 +604,7 @@ def register_ascend_customop(vllm_config: VllmConfig | None = None):
     from vllm.model_executor.custom_op import CustomOp
 
     from vllm_ascend.ops.activation import AscendQuickGELU, AscendSiluAndMul
-    from vllm_ascend.ops.conv import AscendConv2dLayer, AscendConv3dLayer
+    from vllm_ascend.ops.conv import AscendConv3dLayer
     from vllm_ascend.ops.fused_moe.fused_moe import AscendFusedMoE, AscendSharedFusedMoE
     from vllm_ascend.ops.layernorm import AscendGemmaRMSNorm, AscendRMSNorm, AscendRMSNormGated
     from vllm_ascend.ops.linear import (
@@ -616,6 +616,7 @@ def register_ascend_customop(vllm_config: VllmConfig | None = None):
     )
     from vllm_ascend.ops.mla import AscendMultiHeadLatentAttention
     from vllm_ascend.ops.mm_encoder_attention import AscendMMEncoderAttention
+    from vllm_ascend.ops.rel_pos_attention import AscendRelPosAttention
     from vllm_ascend.ops.rotary_embedding import (
         AscendApplyRotaryEmb,
         AscendDeepseekScalingRotaryEmbedding,
@@ -653,8 +654,8 @@ def register_ascend_customop(vllm_config: VllmConfig | None = None):
         "MMEncoderAttention": AscendMMEncoderAttention,
         "ApplyRotaryEmb": AscendApplyRotaryEmb,
         "RMSNormGated": AscendRMSNormGated,
-        "Conv2dLayer": AscendConv2dLayer,
         "Conv3dLayer": AscendConv3dLayer,
+        "RelPosAttention": AscendRelPosAttention,
     }
 
     # 310P: override selected ops with 310P implementations (keep minimal changes outside _310p)