ci: separate benchmark tests into dedicated job

- Exclude benchmarks from the regular test job with -m "not benchmark"
  to prevent GPU OOM from large tensor allocations competing with
  parallel test jobs
- Add a dedicated benchmark job (Python 3.10 only, continue-on-error)
  mirroring the GitLab CI pattern
- Add git safe.directory config to the test job to fix dorny/test-reporter
  failing with git exit code 128 in Docker containers

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Author: camille-004

.github/workflows/ci.yml

@@ -158,6 +158,9 @@ jobs:
       - name: Checkout repository
         uses: actions/checkout@v6
 
+      - name: Mark workspace as safe directory
+        run: git config --global --add safe.directory "$GITHUB_WORKSPACE"
+
       - name: Download wheel
         uses: actions/download-artifact@v7
         with:
@@ -168,7 +171,7 @@ jobs:
         run: pip install dist/*.whl
 
       - name: Run tests
-        run: pytest --durations=10 --junitxml=test-results.xml .
+        run: pytest --durations=10 --junitxml=test-results.xml -m "not benchmark" .
 
       - name: Upload test results
         if: ${{ !cancelled() }}
@@ -185,3 +188,54 @@ jobs:
           name: Test Results (Python ${{ matrix.python-version }})
           path: test-results.xml
           reporter: java-junit
+
+  benchmark:
+    name: Benchmark
+    needs: [images, build]
+    runs-on: [self-hosted, gpu]
+    timeout-minutes: 60
+    continue-on-error: true
+    permissions:
+      contents: read
+      packages: read
+      checks: write
+    container:
+      image: ${{ needs.images.outputs.test_py310 }}
+      credentials:
+        username: ${{ github.actor }}
+        password: ${{ secrets.GITHUB_TOKEN }}
+      options: --gpus all
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v6
+
+      - name: Mark workspace as safe directory
+        run: git config --global --add safe.directory "$GITHUB_WORKSPACE"
+
+      - name: Download wheel
+        uses: actions/download-artifact@v7
+        with:
+          name: wheel-py3.10-linux-x86_64
+          path: dist/
+
+      - name: Install wheel
+        run: pip install dist/*.whl
+
+      - name: Run benchmarks
+        run: pytest --durations=10 --junitxml=benchmark-results.xml -m "benchmark" .
+
+      - name: Upload benchmark results
+        if: ${{ !cancelled() }}
+        uses: actions/upload-artifact@v6
+        with:
+          name: benchmark-results
+          path: benchmark-results.xml
+          retention-days: 7
+
+      - name: Report benchmark results
+        if: ${{ !cancelled() }}
+        uses: dorny/test-reporter@v3
+        with:
+          name: Benchmark Results
+          path: benchmark-results.xml
+          reporter: java-junit