Warn user about PyTorch CPU version

[gtdang]

Have a check on import of the darts theorist. If the user is using the CPU version of torch (autora dependency), it will warn the user that in order to use GPU resources they should upgrade to the GPU version. I'm not sure if there is a way to detect GPU resource availability on the computer. Ideally it would only post the warning if they are using the cpu version of torch AND they have GPU resources.

We can use torch.cuda.is_available() to check if cuda is available. However, I'm not sure if this is a check on the computer or PyTorch. In other words is this a check on if the computer has GPU and the CUDA module? Or is this a check if PyTorch is configured and able to communicate with CUDA.

torch.cuda.device_count() returns the number of GPUs available. I assume CUDA has to be properly set up for torch to return this information though.

To test:

• Build an environment with CPU version of torch on an Oscar instance with GPU. See how these commands function.
• Do the same with the GPU version