Create native function for determining which implementation of SDP to call (#89029)

# Summary
Creates a callable native function that can determine which implementation of scaled dot product will get called. This allows to bump re-order the runtime dispatch of SDP to enable autograd.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/89029
Approved by: https://github.com/cpuhrsch

torchgen/native_function_generation.py

@@ -73,6 +73,7 @@ FUNCTIONAL_OPS_THAT_CANNOT_GET_AN_OUT_VARIANT = [
     "record_stream",  # no return
     "sparse_dim",  # returns an int
     "_nested_tensor_offsets",  # returns a vector of ints
+    "_fused_sdp_choice",  # returns an int
 ]
 
 INPLACE_OPS_THAT_DONT_GET_GROUPED_PROPERLY = [