[inductor] Windows inductor use intel-openmp. (#160258)

After some debug work, I found PyTorch torch_cpu.dll is using intel-openmp, but not MSVC openmp.
So, switch Windows inductor to intel-openmp.

It fixed: c8205cb354/test/inductor/test_aot_inductor.py (L2405-L2408)
<img width="896" height="230" alt="image" src="https://github.com/user-attachments/assets/273b00f8-7dc1-43c9-9b7f-752e16355a80" />

Pull Request resolved: https://github.com/pytorch/pytorch/pull/160258
Approved by: https://github.com/ezyang

setup.py

@@ -1598,6 +1598,7 @@ def main() -> None:
         "networkx>=2.5.1",
         "jinja2",
         "fsspec>=0.8.5",
+        'intel-openmp==2025.1.1 ;platform_system == "Windows" ',  # for Windows inductor
     ]
     if BUILD_PYTHON_ONLY:
         install_requires += [f"{LIBTORCH_PKG_NAME}=={TORCH_VERSION}"]

torch/_inductor/cpp_builder.py

@@ -910,8 +910,15 @@ def _get_python_related_args() -> tuple[list[str], list[str]]:
             str(
                 (
                     Path(sysconfig.get_path("include", scheme="nt")).parent / "libs"
-                ).absolute()
-            )
+                ).absolute()  # python[ver].lib
+            ),
+            str(
+                (
+                    Path(sysconfig.get_path("include", scheme="nt")).parent
+                    / "Library"
+                    / "lib"
+                ).absolute()  # install python librarys location, such as intel-openmp
+            ),
         ]
     else:
         python_lib_path = [sysconfig.get_config_var("LIBDIR")]
@@ -1077,11 +1084,10 @@ def _get_openmp_args(
             libs.append("libiomp5md")
             perload_icx_libomp_win(cpp_compiler)
         else:
-            # /openmp, /openmp:llvm
-            # llvm on Windows, new openmp: https://devblogs.microsoft.com/cppblog/msvc-openmp-update/
-            # msvc openmp: https://learn.microsoft.com/zh-cn/cpp/build/reference/openmp-enable-openmp-2-0-support?view=msvc-170
             cflags.append("openmp")
-            cflags.append("openmp:experimental")  # MSVC CL
+            cflags.append("openmp:experimental")
+            libs.append("libiomp5md")  # intel-openmp
+            ldflags.append("nodefaultlib:vcomp")
     else:
         if config.is_fbcode():
             include_dir_paths.append(build_paths.openmp_include)