[ONNX] Add initial support for FP8 ONNX export (#107962)
This PR resurrects @tcherckez-nvidia's #106379 with changes to resolve conflicts against newer `main` and defines our own constants for the new ONNX types to [avoid breaking Meta's internal usage of an old ONNX](https://github.com/pytorch/pytorch/pull/106379#issuecomment-1675189340).
- `::torch::onnx::TensorProto_DataType_FLOAT8E4M3FN=17`
- `::torch::onnx::TensorProto_DataType_FLOAT8E5M2=19`
Pull Request resolved: https://github.com/pytorch/pytorch/pull/107962
Approved by: https://github.com/justinchuby, https://github.com/titaiwangms