Add channel conda-forge to environment.yml for Windows support

[SetZero]

Currently the nvidia channel is missing cudatoolkit in version 11.1 for Windows. (https://anaconda.org/nvidia/cudatoolkit/files)
Channel conda-forge has these files. So I'd suggest to add this channel to the environment.yml for Windows support.

Output on Windows:

foo@bar:~$ conda env create -f environment.yml

Collecting package metadata (repodata.json): done
Solving environment: failed

ResolvePackageNotFound:
  - cudatoolkit=11.1

[tljstewart]

@SetZero but its interesting, because there readme says that a custom cuda kernel is used from Nvidia... does this change work for you?

Why is CUDA toolkit installation necessary?
The PyTorch package contains the required CUDA toolkit libraries needed to run PyTorch, so why is a separate CUDA toolkit installation required? Our models use custom CUDA kernels to implement operations such as efficient resampling of 2D images. PyTorch code invokes the CUDA compiler at run-time to compile these kernels on first-use. The tools and libraries required for this compilation are not bundled in PyTorch and thus a host CUDA toolkit installation is required.

https://github.com/NVlabs/stylegan3/blob/main/docs/troubleshooting.md#why-is-cuda-toolkit-installation-necessary