Update on "move has_torch_function to C++, and make a special case ob…

…ject_has_torch_function"


This PR pulls `__torch_function__` checking entirely into C++, and adds a special `object_has_torch_function` method for ops which only have one arg as this lets us skip tuple construction and unpacking. We can now also do away with the Python side fast bailout for `Tensor` (e.g. `if any(type(t) is not Tensor for t in tensors) and has_torch_function(tensors)`) because they're actually slower than checking with the Python C API.

Test plan: Existing unit tests. Benchmarks are in #48966

Differential Revision: [D25590732](https://our.internmc.facebook.com/intern/diff/D25590732)

[ghstack-poisoned]

torch/nn/functional.py

@@ -1711,7 +1711,7 @@ def log_softmax(input: Tensor, dim: Optional[int] = None, _stacklevel: int = 3,
           is performed. This is useful for preventing data type overflows. Default: None.
     """
     if not torch.jit.is_scripting():
-        if has_torch_function_unary(logits):
+        if has_torch_function_unary(input):
             return handle_torch_function(log_softmax, (input,), input, dim=dim, _stacklevel=_stacklevel, dtype=dtype)
     if dim is None:
         dim = _get_softmax_dim("log_softmax", input.dim(), _stacklevel)