[Attention] Refactor FA block_size limitations to hybrid models only (#29084)

Signed-off-by: NickLucche <nlucches@redhat.com>

tests/v1/worker/test_gpu_model_runner.py

@@ -185,7 +185,9 @@ def _make_mock_backend_for_kernel_block_size(
     supported_sizes: list[int | MultipleOf],
 ):
     class _MockBackend:
-        supported_kernel_block_sizes = supported_sizes
+        @staticmethod
+        def get_supported_kernel_block_sizes():
+            return supported_sizes
 
     return _MockBackend()