CI: pin numpy<2.5 (mlir) and pytest<9.1 (cuda-core released tests)

Two nightly failure fixups after the first green iteration:

nightly-numba-cuda-mlir: numba-cuda-mlir 0.4.0 has an inverted guard
that registers an overload of np.row_stack on NumPy 2.x, and NumPy 2.5
removed that name entirely, so test collection fails with
"AttributeError: module 'numpy' has no attribute 'row_stack'". Cap
numpy to <2.5. See NVIDIA/numba-cuda-mlir#154.

nightly-cuda-core: released cuda-core v1.0.1's test suite uses a
parametrize argvalues pattern that pytest 9.1 rejects
("in parametrize the number of names (1)... must be equal to the
number of values (3)"). The main-side fix was #2212 but it has not
shipped in a cuda-core release yet. Cap pytest to <9.1 for the
released-cuda-core test run only.

Author: leofang

.github/workflows/test-wheel-linux.yml

@@ -475,5 +475,9 @@ jobs:
           # that cuda-core version shipped with.
           pip install --upgrade "pip>=25.1"
           pip install --group "${CUDA_CORE_TEST_GROUP}"
+          # Cap pytest below 9.1: released cuda-core <=1.0.1 has parametrize
+          # patterns that pytest 9.1 rejects; the main-side fix (#2212) has
+          # not yet shipped in a cuda-core release.
+          pip install "pytest<9.1"
           pytest -rxXs -v --durations=0 --randomly-dont-reorganize tests/
           popd

.github/workflows/test-wheel-windows.yml

@@ -456,5 +456,8 @@ jobs:
           pushd cuda-core-released/cuda_core
           pip install --upgrade "pip>=25.1"
           pip install --group "${CUDA_CORE_TEST_GROUP}"
+          # Cap pytest below 9.1 — released cuda-core <=1.0.1 has parametrize
+          # patterns that pytest 9.1 rejects (see #2212).
+          pip install "pytest<9.1"
           pytest -rxXs -v --durations=0 --randomly-dont-reorganize tests/
           popd

ci/tools/run-tests

@@ -163,7 +163,9 @@ elif [[ "${test_module}" == "core" || "${test_module}" == nightly-* ]]; then
       )
     elif [[ "${test_module}" == "nightly-numba-cuda-mlir" ]]; then
       echo "Installing pathfinder + bindings + core + numba-cuda-mlir"
-      PIP_ARGS+=("numba-cuda-mlir[cu${TEST_CUDA_MAJOR}]")
+      # numpy<2.5: numba-cuda-mlir 0.4.0 registers np.row_stack, which was
+      # removed in NumPy 2.5. See NVIDIA/numba-cuda-mlir#154.
+      PIP_ARGS+=("numba-cuda-mlir[cu${TEST_CUDA_MAJOR}]" "numpy<2.5")
     fi
 
     pip install "${PIP_ARGS[@]}"