pyi codegen refactor - no need to group python signatures by overload name (#49057)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/49057
Now that all of the byte-for-byte hacks are removed in the pyi codegen, there's no reason for the codegen to group pyi signature overloads together. I updated the logic in `gen_pyi` that computes signatures (`generate_type_hints()` and _generate_named_tuples()`) to operate per individual `PythonSignatureGroup`
Test Plan: Imported from OSS
Reviewed By: ezyang
Differential Revision: D25410849
Pulled By: bdhirsh
fbshipit-source-id: 8c190035d7bfc06ed192468efbe7d902922ad1fa
