[1/N][CustomOp] Register activation customop instead of overwrite forward_oot (#1841)

### What this PR does / why we need it?
We'll refator `CustomOp` in vllm-ascend from this pr on. 

Use function `CustomOp.register_oot` to achieve the customop registery,
taking `AscendQuickGELU` as an example:
```python
from vllm_ascend.ops.activation import AscendQuickGELU
CustomOp.register_oot(_decorated_op_cls=AscendQuickGELU, name="QuickGELU")
```

This is a quick adapt for `CustomOp.register_oot` mechanism from vllm
0.9.2. For further step, we can remove inherit from `QuickGELU` can
write our own `QuickGELU` at all.

Part of https://github.com/vllm-project/vllm-ascend/pull/1647



- vLLM version: v0.9.2
- vLLM main:
8dfb45ca33

---------

Signed-off-by: MengqingCao <cmq0113@163.com>

tests/e2e/singlecard/test_aclgraph.py

@@ -36,7 +36,7 @@ MODELS = [
 
 @pytest.mark.parametrize("model", MODELS)
 @pytest.mark.parametrize("max_tokens", [32])
-def test_models(
+def test_models_with_aclgraph(
     model: str,
     max_tokens: int,
 ) -> None:
@@ -48,12 +48,12 @@ def test_models(
     sampling_params = SamplingParams(max_tokens=max_tokens, temperature=0.0)
     # TODO: change to use vllmrunner when the registry of custom op is solved
     # while running pytest
-    vllm_model = LLM(model)
+    vllm_model = LLM(model, max_model_len=1024)
     vllm_aclgraph_outputs = vllm_model.generate(prompts, sampling_params)
     del vllm_model
     torch.npu.empty_cache()
 
-    vllm_model = LLM(model, enforce_eager=True)
+    vllm_model = LLM(model, enforce_eager=True, max_model_len=1024)
     vllm_eager_outputs = vllm_model.generate(prompts, sampling_params)
     del vllm_model
     torch.npu.empty_cache()