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/extension/test_backend_equivalence.py

@@ -12,7 +12,8 @@ from astrai.inference.cache import PagePool, TaskCacheManager
 from astrai.inference.runtime.graph import CudaGraphContext
 from astrai.inference.scheduler import InferenceScheduler
 from astrai.inference.workspace import InferenceWorkspace
-from tests.extension.conftest import D, skip_no_kernel
+from tests.conftest import skip_no_kernel
+from tests.extension.conftest import D
 from tests.helpers import FakeTokenizer