Support for PyTorch Lightning

[jeremyfowers]

Problem Statement

TurnkeyML currently supports only torch.nn.modules within the turnkey benchmark command. However, there is interest in using TurnkeyML with PyTorch Lightning modules as well.

Scope of Work

At least the following improvements would be needed to support PyTorch Lightning:

1. In the analyze module, detect PyTorch Lightning modules and pass them to explore_invocation(). Currently, we only pass torch.nn.module to explore_invocation().
2. Add a new build Stage to export the Lightning module to ONNX. This is hopefully as simple as calling lightning_module.to_onnx().
3. Add this new build stage to the ExportPlaceholder Stage so that it can be automatically selected whenever a Lightning module is encountered.

Contribution

• I am setting a priority of P2 until we have an e2e example available that we can use for testing.
• I believe that this would make a good first issue if someone wanted to contribute it.

cc @andife with regards to #69