[TEST] Skip test_schema_correctness for float8 dtype (pytorch#115757)

According to the pytorch#107256 (comment) the ops tested in `test_schema_correctness` are not supported with `torch.float8_e4m3fn` yet. Until they are not supported, it is best to skip the test.

Pull Request resolved: pytorch#115757
Approved by: https://github.com/drisspg

torch/testing/_internal/common_methods_invocations.py

@@ -14228,6 +14228,10 @@ def reference_flatten(input, start_dim=0, end_dim=-1):
             # Sample inputs isn't really parametrized on dtype
             DecorateInfo(unittest.skip("Skipped!"), 'TestCommon', 'test_dtypes',
                          device_type='cuda'),
+            # "mul_cuda" not implemented for float8_e4m3fn
+            # https://github.com/pytorch/pytorch/issues/107256
+            DecorateInfo(unittest.skip("Skipped!"), 'TestSchemaCheckModeOpInfo', 'test_schema_correctness',
+                         dtypes=(torch.float8_e4m3fn,)),
         )
     ),
     OpInfo(