[optim][radam] default to foreach when CUDA + differentiable=False

[janeyx99]

Stack from ghstack (oldest at bottom):

• [optim][sgd] default to foreach when CUDA + differentiable=False #92730
• [optim][rprop] default to foreach when CUDA + differentiable=False #92728
• [optim][rmsprop] default to foreach when CUDA + differentiable=False #92727
• -> [optim][radam] default to foreach when CUDA + differentiable=False #92726
• [optim][asgd] default to foreach when CUDA + differentiable=False #92724
• [optim][adamax] default to foreach when CUDA + differentiable=False #92723
• [optim][adagrad] default to foreach when CUDA + differentiable=False #92716

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/92726

• 📄 Preview Python docs built from this PR
• 📄 Preview C++ docs built from this PR
• ❓ Need help or want to give feedback on the CI? Visit the bot commands wiki or our office hours

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit 21360d2:
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[albanD]

Good!