fix(tests): disable GPU on macOS CI to avoid Metal paravirtual device timeout

[Copilot]

Integration tests performing text generation were timing out after 5 minutes on macOS CI runners. The virtualized GPU (Apple Paravirtual device) causes Metal library initialization delays (11+ seconds) and text generation hangs.

Changes

• Disable GPU layers on macOS (gpuLayers: 0) for integration tests that perform text generation
• Other platforms continue using GPU acceleration (gpuLayers: 99)
• Applied to 8 test cases across 3 files:
  • test/chat_session_test.dart (1 instance)
  • test/engine_integration_test.dart (4 instances)
  • test/grammar_regression_test.dart (3 instances)

Example

// Disable GPU on macOS CI to avoid Metal paravirtual device issues
final gpuLayers = Platform.isMacOS ? 0 : 99;

await engine.loadModel(
  modelFile.path,
  modelParams: ModelParams(
    contextSize: 256,
    logLevel: LlamaLogLevel.none,
    gpuLayers: gpuLayers,
  ),
);

This forces CPU-only inference on macOS while maintaining GPU acceleration on Linux/Windows where hardware GPU access is reliable.

Original prompt

investigate https://github.com/leehack/llamadart/actions/runs/21601294578/job/62247815048#step:5:1

why it's failing only macos?

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 47.18%. Comparing base (b6fcad5) to head (4234d6b).

Additional details and impacted files

@@           Coverage Diff           @@
##             main      #31   +/-   ##
=======================================
  Coverage   47.18%   47.18%           
=======================================
  Files          23       23           
  Lines        1757     1757           
=======================================
  Hits          829      829           
  Misses        928      928           

Flag	Coverage Δ
macos-latest	43.56% <0.00%> (+0.25%)	⬆️
ubuntu-latest	43.24% <0.00%> (ø)
web	80.11% <0.00%> (ø)
windows-latest	43.24% <0.00%> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.