Support TorchScript Modules

[sbrunk]

We want to be able to load and use TorchScript modules. For inference but ideally also for training and for using a TorchScript modules as part of a regular module.

The PyTorch C++ frontend already supports loading serialized TorchScript models and it is already mapped in the JavaCPP preset. See https://github.com/bytedeco/javacpp-presets/blob/master/pytorch/samples/ExampleApp.java for an example