Allow flexible returns from a forward function.

[maxwest-uw]

Change Description

resolves #858

hyrax users can now return a dict of torch.Tensors from a model's forward function.

Solution Description

• change is backwards compatible (i.e. user can return either a dict or a tensor from the forward function and everything should still work)
• in cases where the forward function just returns a tensor, the data will eventually be converted into a dict with the format { "data": <tensor> }, which is how the data will eventually be saved to lance db anyway.
• ResultDataset should now always return a dict with the data for a give index or set of indices
• in cases throughout the data processing pipeline where we assume that a result is going to be a tensor, we have added a new function ResultDataset.get_combined_tensor, which for a given index or list of indices will return a concatenated and flattened tensor of all the available data keys.

Code Quality

• I have read the Contribution Guide and agree to the Code of Conduct
• My code follows the code style of this project
• My code builds (or compiles) cleanly without any errors or warnings
• My code contains relevant comments and necessary documentation