[Kernel] add custom op GmmSwigluQuantWeightNzTensorList (#3804)

### What this PR does / why we need it?

This PR introduces support for adding custom CANN `aclnn` ops to
`vllm-ascend`, allowing users to define and use their own custom
operators.

Key changes include:
- Building and installing custom ops into the `vllm-ascend`-specified
directory
- Binding the `aclnn` op interface to the `torch.ops._C_ascend` module
- Enabling invocation of these ops within `vllm-ascend`

This PR includes a sample custom op:
`aclnnGroupedMatmulSwigluQuantWeightNzTensorList`, which is adapted from
the CANN operator
[`aclnnGroupedMatmulSwigluQuantWeightNZ`](https://www.hiascend.com/document/detail/zh/canncommercial/83RC1/API/aolapi/context/aclnnGroupedMatmulSwigluQuantWeightNZ.md).
Its input parameters `weight` and `weight_scale` now accept
`list[torch.Tensor]` (i.e., `at::TensorList`).

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

No.


- vLLM version: v0.11.2

---------

Signed-off-by: QianChenxi <chenxi.qian.cq@outlook.com>

setup.py

@@ -25,7 +25,7 @@ import sys
 from sysconfig import get_paths
 from typing import Dict, List
 
-from setuptools import Extension, find_packages, setup
+from setuptools import Command, Extension, find_packages, setup
 from setuptools.command.build_ext import build_ext
 from setuptools.command.build_py import build_py
 from setuptools.command.develop import develop
@@ -199,6 +199,27 @@ class custom_build_info(build_py):
         super().run()
 
 
+class build_and_install_aclnn(Command):
+    description = "Build and install AclNN by running build_aclnn.sh"
+    user_options = []
+
+    def initialize_options(self):
+        pass
+
+    def finalize_options(self):
+        pass
+
+    def run(self):
+        try:
+            print("Running bash build_aclnn.sh ...")
+            subprocess.check_call(
+                ["bash", "csrc/build_aclnn.sh", ROOT_DIR, envs.SOC_VERSION])
+            print("buid_aclnn.sh executed successfully!")
+        except subprocess.CalledProcessError as e:
+            print(f"Error running build_aclnn.sh: {e}")
+            raise SystemExit(e.returncode)
+
+
 class cmake_build_ext(build_ext):
     # A dict of extension directories that have been configured.
     did_config: Dict[str, bool] = {}
@@ -385,8 +406,22 @@ class cmake_build_ext(build_ext):
                         shutil.copy(src_path, dst_path)
                         print(f"Copy: {src_path} -> {dst_path}")
 
+        # copy back _cann_ops_custom directory
+        src_cann_ops_custom = os.path.join(ROOT_DIR, "vllm_ascend",
+                                           "_cann_ops_custom")
+        dst_cann_ops_custom = os.path.join(self.build_lib, "vllm_ascend",
+                                           "_cann_ops_custom")
+        if os.path.exists(src_cann_ops_custom):
+            import shutil
+            if os.path.exists(dst_cann_ops_custom):
+                shutil.rmtree(dst_cann_ops_custom)
+            shutil.copytree(src_cann_ops_custom, dst_cann_ops_custom)
+            print(f"Copy: {src_cann_ops_custom} -> {dst_cann_ops_custom}")
+
     def run(self):
-        # First, run the standard build_ext command to compile the extensions
+        # First, ensure ACLNN custom-ops is built and installed.
+        self.run_command("build_aclnn")
+        # Then, run the standard build_ext command to compile the extensions
         super().run()
 
 
@@ -450,6 +485,7 @@ def get_requirements() -> List[str]:
 cmdclass = {
     "develop": custom_develop,
     "build_py": custom_build_info,
+    "build_aclnn": build_and_install_aclnn,
     "build_ext": cmake_build_ext,
     "install": custom_install
 }