refactor: unify CUDA skip guards in tests

- add skip_no_fp8 (CUDA + fp8_mm kernel + cc 8.9+) to tests/conftest.py
- use skip_no_cuda / skip_no_kernel / skip_no_fp8 directly in test modules
- drop _GPU alias and tests.extension.conftest re-exports
- remove unused imports (Union in hf_adapter, make_grpo_config in data conftest)

tests/conftest.py

@@ -15,8 +15,17 @@ from tests.helpers import (
 
 CUDA_AVAIL = torch.cuda.is_available()
 KERNEL_AVAIL = CUDA_AVAIL and all(is_available(k) for k in KERNEL_NAMES)
+FP8_AVAIL = (
+    CUDA_AVAIL
+    and is_available("fp8_mm")
+    and torch.cuda.get_device_capability() >= (8, 9)
+)
 skip_no_cuda = pytest.mark.skipif(not CUDA_AVAIL, reason="CUDA not available")
 skip_no_kernel = pytest.mark.skipif(not KERNEL_AVAIL, reason="CUDA kernels not built")
+skip_no_fp8 = pytest.mark.skipif(
+    not FP8_AVAIL,
+    reason="fused FP8 MMA requires a built kernel and compute capability 8.9+",
+)
 
 
 def pytest_configure(config):