docs(glm): add GLM-5.3 training coverage

[HuiyingLi]

Summary

• Extend the existing GLM MoE/DSA model coverage page through GLM-5.3.
• Document the EP64/PP4 recipe, cuDNN DSA + HybridEP path, expert resharding setting, launch command, Hugging Face logits parity, and training results.
• Add GLM-5.3 to the README news and supported-model table.
• Update the nightly documentation navigation label.

This follows the structure of #3744 and documents the support merged in #3747.

Validation

• MDX syntax validation: passed
• Fern configuration check: 0 errors
• Filesystem model-coverage checks: 4 passed
• git diff --check: passed

Training evidence

• First-four-layer Hugging Face parity: 0.639% logits relative L2 error and 96.61% top-1 agreement.
• Full 32-node / 256-GPU run: 100/100 steps completed; loss 1.33347 at step 0 and 0.53756 at step 99.
• Weights & Biases run

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[github-actions]

🌿 Preview your docs: https://nvidia-preview-preview-ae73159cebf3.docs.buildwithfern.com/nemo/automodel

[HuiyingLi]

/ok to test 8fe3d33

[HuiyingLi]

/ok to test ae73159