build: Use no-build-isolation to install deep_gemm to fix arm install

[chtruong814]

What does this PR do ?

Use no-build-isolation to install deep_gemm to fix arm install

Otherwise, the install fails with a stack trace like this. Inside an arm container, I was able to install deep_gemm with uv pip install --no-build-isolation

Using Python 3.12.3 environment at: /opt/nemo_rl_venv
    Updated https://github.com/deepseek-ai/DeepGEMM.git (7b6b5563b9d4c1ae07ffbce7f78ad3ac9204827c)
  × Failed to build `deep-gemm @ git+https://github.com/deepseek-ai/DeepGEMM.git@7b6b5563b9d4c1ae07ffbce7f78ad3ac9204827c`
  ├─▶ The build backend returned an error
  ╰─▶ Call to `setuptools.build_meta.build_wheel` failed (exit status: 1)

      [stderr]
      /opt/uv_cache/builds-v0/.tmpSkF3nO/lib/python3.12/site-packages/torch/_subclasses/functional_tensor.py:279:
      UserWarning: Failed to initialize NumPy: No module named 'numpy' (Triggered internally at
      /pytorch/torch/csrc/utils/tensor_numpy.cpp:81.)
        cpu = _conversion_method_template(device=torch.device("cpu"))
      Traceback (most recent call last):
        File "<string>", line 14, in <module>
        File "/opt/uv_cache/builds-v0/.tmpSkF3nO/lib/python3.12/site-packages/setuptools/build_meta.py", line 331, in
      get_requires_for_build_wheel
          return self._get_build_requires(config_settings, requirements=[])
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
        File "/opt/uv_cache/builds-v0/.tmpSkF3nO/lib/python3.12/site-packages/setuptools/build_meta.py", line 301, in
      _get_build_requires
          self.run_setup()
        File "/opt/uv_cache/builds-v0/.tmpSkF3nO/lib/python3.12/site-packages/setuptools/build_meta.py", line 317, in
      run_setup
          exec(code, locals())
        File "<string>", line 91, in <module>
        File "/opt/uv_cache/builds-v0/.tmpSkF3nO/lib/python3.12/site-packages/torch/utils/cpp_extension.py", line 1337,
      in CUDAExtension
          library_dirs += library_paths(device_type="cuda")
                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
        File "/opt/uv_cache/builds-v0/.tmpSkF3nO/lib/python3.12/site-packages/torch/utils/cpp_extension.py", line 1548,
      in library_paths
          if (not os.path.exists(_join_cuda_home(lib_dir)) and
                                 ^^^^^^^^^^^^^^^^^^^^^^^^
        File "/opt/uv_cache/builds-v0/.tmpSkF3nO/lib/python3.12/site-packages/torch/utils/cpp_extension.py", line 2982,
      in _join_cuda_home
          raise OSError('CUDA_HOME environment variable is not set. '
      OSError: CUDA_HOME environment variable is not set. Please set it to your CUDA install root.

      hint: This usually indicates a problem with the package or the build environment.

Also, it seems that a PR #859 merged into main that should not have made it through the CI. A Github error happened where the build was skipped, and the CI quality check should have failed but did not. I updated the check to verify the build succeeded when tests are ran and also had to increase the allowed GPU hours for the nightly test suite.

Merge queue that should not have passed:
https://github.com/NVIDIA-NeMo/RL/actions/runs/17169685093/job/48716856164

Issues

List issues that this PR closes (syntax):

Usage

• You can potentially add a usage example below

# Add a code snippet demonstrating how to use this

Before your PR is "Ready for review"

Pre checks:

• Make sure you read and followed Contributor guidelines
• Did you write any new necessary tests?
• Did you run the unit tests and functional tests locally? Visit our Testing Guide for how to run tests
• Did you add or update any necessary documentation? Visit our Document Development Guide for how to write, build and test the docs.

Additional Information

• ...