"Fix" PyTorch CPU conda testing

It's still horribly broken, but make it a bit better by not installing
pytorch from default anaconda channel (which installs 1.12.1 that does
not have any dependencies 2.0 dev package supposed to have)

For example, see this runlog
https://github.com/pytorch/pytorch/actions/runs/4155371267/jobs/7189101147

Author: malfet

windows/internal/smoke_test.bat

@@ -64,7 +64,7 @@ echo "install conda package"
 set "CONDA_HOME=%CD%\conda"
 set "tmp_conda=%CONDA_HOME%"
 set "miniconda_exe=%CD%\miniconda.exe"
-set "CONDA_EXTRA_ARGS="
+set "CONDA_EXTRA_ARGS=cpuonly -c pytorch-nightly"
 if "%CUDA_VERSION%" == "116" (
     set "CONDA_EXTRA_ARGS=pytorch-cuda=11.6 -c nvidia -c pytorch-nightly"
 )
@@ -102,15 +102,6 @@ set CUDA_VERSION_STR=%CUDA_VER_MAJOR%.%CUDA_VER_MINOR%
 for /F "delims=" %%i in ('where /R "%PYTORCH_FINAL_PACKAGE_DIR:/=\%" *.tar.bz2') do call conda install -yq "%%i" --offline
 if ERRORLEVEL 1 exit /b 1
 
-if "%CUDA_VERSION%" == "cpu" goto install_cpu_torch
-
-goto smoke_test
-
-:install_cpu_torch
-call conda install %CONDA_EXTRA_ARGS% -y cpuonly -c pytorch-nightly
-if ERRORLEVEL 1 exit /b 1
-
-:smoke_test
 python -c "import torch"
 if ERRORLEVEL 1 exit /b 1