Is your feature request related to a problem? Please describe.

If a user follows the current guide to install MONAI, the system will end up using pytorch cpu version, even if user's system has NVIDIA GPUs correctly installed and configured. Adding Windows GPU support will expand the user base and MONAI will benefit from the increased diversity from different platforms.

Example environment:

• Windows 10
• Conda
• NVIDIA RTX 3070, Driver 512.36, CUDA 11.3
• CUDA_PATH is set in system paths

Both of these instructions will end up using PyTorch-cpu

pip install monai

conda install monai -c corda-forge

Describe the solution you'd like

The installation guide should include method to install MONAI on Windows NVIDIA-GPU.
For example, this is working for me:

1. Install NVIDIA latest driver
2. Check PyTorch Official Guide for the recommend CUDA versions. For Pip package, the user needs to download the CUDA manually, install it on the system, and ensure CUDA_PATH is set properly.
3. Continue to follow the guide and install PyTorch. For example for CUDA 11.3, one needs to run

pip3 install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu113

4. Install MONAI using Pip (or Conda, I didn't try though)

Describe alternatives you've considered

At least, a link to the official installation of PyTorch will help Windows user.

Additional context