[AO] Clear the contents of the torch/ao/__init__.py (#69415)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/69415
Adding the imports inside the torch/ao/__init__.py has a high chance of causing circular dependencies, especially if sparsity and quantization use each other's resources.
To avoid the dependency issues, we can just keep the __init__ empty.
Notes:
- This means that the user will have to explicitly import the `torch.ao.quantization` or `torch.ao.sparsity` instead of `from torch import ao; ao.quantization.???`.
- The issue of circular dependencies that are caused by the imports with binding submodules is [fixed in Python 3.7](https://docs.python.org/3/whatsnew/3.7.html#other-language-changes), which means this solution will become obsolete at the [3.6's EoL](https://www.python.org/dev/peps/pep-0494/#and-beyond-schedule), which comes [12/23/2022](https://devguide.python.org/#status-of-python-branches).
Future options to resolve the circular dependencies (subject to discussion):
1. Use interfaces for binding submodules. For example, have a torch/ao/_nn with all the source code, and an interface torch/ao/nn with only the __init__.py file. The __init__ files inside the torch/ao/_nn will be empty
2. Completely isolate the common code into a separate submodule, s.a. torch/ao/common. The other submodules will not be referencing each other.
Test Plan: Imported from OSS
Reviewed By: vkuzo
Differential Revision: D32860168
Pulled By: z-a-f
fbshipit-source-id: e3fe77e285992d34c87d8742e1a5e449ce417c36