Enable Arm Hosted GitHub Action Runners for ggml-org/llama.cpp and ggml-org/whisper.cpp

Link to your GitHub repository

• https://github.com/ggml-org/llama.cpp
• https://github.com/ggml-org/whisper.cpp

Please describe your usecase

We want to enable continuous integration coverage on Arm Graviton4 (c8g) instances as part of KFI-111. Both llama.cpp and whisper.cpp are high-visibility open-source ML projects, and adding Graviton4 runners will ensure build correctness, detect regressions early, and improve Arm ecosystem visibility.
The workflows will be lightweight: build jobs plus short functional smoke tests (e.g., CLI --help, llama-bench help). Longer benchmarks will remain internal only.

Estimate on Runner usage - On monthly basis, how many hours you expect to run the runners.

Estimated 30–50 hours per month total across both repos.
Typical usage: 1–2 jobs per PR, runtime 10–15 minutes each. Concurrency needs are medium (8-16 parallel jobs).