Add patch to gcc 10 specific to nccl

[hmaarrfk]

Checklist

• [ ] Used a personal fork of the feedstock to propose changes
• [ ] Bumped the build number (if the version is unchanged)
• [ ] Reset the build number to 0 (if the version changed)
• [ ] Re-rendered with the latest conda-smithy (Use the phrase code>@<space/conda-forge-admin, please rerender in a comment in this PR for automated rerendering)
• [ ] Ensured the license file is being packaged.

[conda-forge-linter]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

[hmaarrfk]

@xhochy i believe that this should help with compatibility with certain cuda errors.

[hmaarrfk]

I uploaded the linux-64 builds for linux-64 on my channel, and we can test the results with cuda 11.1 and 11.2 https://github.com/conda-forge/pytorch-cpu-feedstock/pull/95

[hmaarrfk]

I was able to successfully build pytorch 1.11 with this patch:

# To have conda build upload to anaconda.org automatically, use
# conda config --set anaconda_upload yes
anaconda upload \
    /home/conda/feedstock_root/build_artifacts/linux-64/pytorch-1.11.0-cuda112py39hda16ee4_1.tar.bz2 \
    /home/conda/feedstock_root/build_artifacts/linux-64/pytorch-gpu-1.11.0-cuda112py39h0bbbad9_1.tar.bz2
anaconda_upload is not set.  Not uploading wheels: []

INFO :: The inputs making up the hashes for the built packages are as follows:
{
  "pytorch-1.11.0-cuda112py39hda16ee4_1": {
    "recipe": {
      "c_compiler": "gcc",
      "c_compiler_version": "10",
      "channel_targets": "conda-forge main",
      "cuda_compiler": "nvcc",
      "cuda_compiler_version": "11.2",
      "cudnn": "8",
      "cxx_compiler": "gxx",
      "cxx_compiler_version": "10",
      "libprotobuf": "3.19",
      "mkl": "2022",
      "nccl": "2",
      "numpy": "1.19",
      "python": "3.9.* *_cpython",
      "target_platform": "linux-64"
    }
  },
  "pytorch-gpu-1.11.0-cuda112py39h0bbbad9_1": {
    "recipe": {
      "channel_targets": "conda-forge main",
      "cuda_compiler_version": "11.2",
      "target_platform": "linux-64"
    }
  }
}

I didn't make the logs. but hopefully this builds confidence that we should include this.

[hmaarrfk]

Thank you for the review!