Add Python 3.15 to the build/test matrix

Backport of #2108 to the 12.9.x branch so that cuda-bindings 12.9.x
wheels are available for Python 3.15, which the main branch needs in
order to build cuda-core.

- Add 3.15 / 3.15t to the build matrix
- Upgrade cibuildwheel to v4.0.0rc1 (required for cpython-prerelease)
- Enable cpython-prerelease in CIBW_ENABLE
- Add amd64-only test entries for 3.15 / 3.15t (CUDA 12.9.1)

Author: leofang

.github/workflows/build-wheel.yml

@@ -32,6 +32,8 @@ jobs:
           - "3.13t"
           - "3.14"
           - "3.14t"
+          - "3.15"
+          - "3.15t"
     name: py${{ matrix.python-version }}
     runs-on: ${{ (inputs.host-platform == 'linux-64' && 'linux-amd64-cpu8') ||
                  (inputs.host-platform == 'linux-aarch64' && 'linux-arm64-cpu8') ||
@@ -111,7 +113,7 @@ jobs:
           if-no-files-found: error
 
       - name: Build cuda.core wheel
-        uses: pypa/cibuildwheel@8d2b08b68458a16aeb24b64e68a09ab1c8e82084  # v3.4.1
+        uses: pypa/cibuildwheel@54327ab9d35de03b359ac25c97de9417d94639c0  # v4.0.0rc1
         env:
           CIBW_BUILD: ${{ env.CIBW_BUILD }}
           CIBW_ARCHS_LINUX: "native"
@@ -120,7 +122,7 @@ jobs:
           CIBW_REPAIR_WHEEL_COMMAND_WINDOWS: "delvewheel repair --namespace-pkg cuda -w {dest_dir} {wheel}"
           CIBW_ENVIRONMENT: >
             CUDA_PYTHON_PARALLEL_LEVEL=${{ env.CUDA_PYTHON_PARALLEL_LEVEL }}
-          CIBW_ENABLE: "cpython-freethreading"
+          CIBW_ENABLE: "cpython-freethreading cpython-prerelease"
         with:
           package-dir: ./cuda_core/
           output-dir: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
@@ -154,7 +156,7 @@ jobs:
           cuda-version: ${{ inputs.cuda-version }}
 
       - name: Build cuda.bindings wheel
-        uses: pypa/cibuildwheel@8d2b08b68458a16aeb24b64e68a09ab1c8e82084  # v3.4.1
+        uses: pypa/cibuildwheel@54327ab9d35de03b359ac25c97de9417d94639c0  # v4.0.0rc1
         env:
           CIBW_BUILD: ${{ env.CIBW_BUILD }}
           CIBW_ARCHS_LINUX: "native"
@@ -168,7 +170,7 @@ jobs:
             CUDA_PYTHON_PARALLEL_LEVEL=${{ env.CUDA_PYTHON_PARALLEL_LEVEL }}
           CIBW_BEFORE_BUILD_WINDOWS: "pip install delvewheel"
           CIBW_REPAIR_WHEEL_COMMAND_WINDOWS: "delvewheel repair --namespace-pkg cuda -w {dest_dir} {wheel}"
-          CIBW_ENABLE: "cpython-freethreading"
+          CIBW_ENABLE: "cpython-freethreading cpython-prerelease"
         with:
           package-dir: ./cuda_bindings/
           output-dir: ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}

ci/test-matrix.json

@@ -15,6 +15,8 @@
       { "ARCH": "amd64", "PY_VER": "3.13t", "CUDA_VER": "12.9.1", "LOCAL_CTK": "1", "GPU": "l4", "DRIVER": "latest" },
       { "ARCH": "amd64", "PY_VER": "3.14", "CUDA_VER": "12.9.1", "LOCAL_CTK": "1", "GPU": "l4", "DRIVER": "latest" },
       { "ARCH": "amd64", "PY_VER": "3.14t", "CUDA_VER": "12.9.1", "LOCAL_CTK": "0", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.15",  "CUDA_VER": "12.9.1", "LOCAL_CTK": "0", "GPU": "l4", "DRIVER": "latest" },
+      { "ARCH": "amd64", "PY_VER": "3.15t", "CUDA_VER": "12.9.1", "LOCAL_CTK": "0", "GPU": "l4", "DRIVER": "latest" },
       { "ARCH": "arm64", "PY_VER": "3.10", "CUDA_VER": "12.9.1", "LOCAL_CTK": "1", "GPU": "a100", "DRIVER": "latest" },
       { "ARCH": "arm64", "PY_VER": "3.10", "CUDA_VER": "13.0.1", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest" },
       { "ARCH": "arm64", "PY_VER": "3.11", "CUDA_VER": "12.9.1", "LOCAL_CTK": "0", "GPU": "a100", "DRIVER": "latest" },