This PR establishes a new nightly workflow via GitHub actions which will run our unit test suite. The workflow will be triggered at 02:15 UTC daily.

Builds are completed for these platforms using standard GitHub runners:

• Linux x86-64
  • CPU
  • CUDA 11.8.0, targeting SM75
  • CUDA 12.8.0, targeting SM75
• Windows x86-64
  • CPU
  • CUDA 11.8.0, targeting SM75
  • CUDA 12.8.1, targeting SM75

Unit tests are executed with our minimum Python version requirement of 3.9.

• Linux x86-64
  • CPU: PyTorch 2.7.0+cpu
  • CUDA 11.8.0: PyTorch 2.4.1+cu118 on NVIDIA T4
  • CUDA 12.8.1: PyTorch 2.7.0+cu128 on NVIDIA T4
• Windows x86-64
  • CPU: PyTorch 2.7.0+cpu
  • CUDA 11.8.0: PyTorch 2.4.1+cu118 on NVIDIA T4

Minor adjustments to the test suite are made to limit the runtime of the CPU tests. The paged optimizer tests will be skipped for now on Windows as well.