RTen models should have provenance metadata

[robertknight]

RTen models should have machine-readable metadata in them that specifies where they came from. At a minimum this would include:

• Filename and hash (SHA-1 ?) of the ONNX model that was used to generate it
• Date when the RTen model was created

This enables verifying that a .rten model corresponds to a particular ONNX model. The rten-cli package could support a --metadata flag that would dump this info from a compiled model.

In addition it would be useful to also incorporate metadata from the ONNX model. See https://github.com/onnx/onnx/blob/main/docs/IR.md#models. In particular I would like to encourage including:

• The URL of the associated repositories
• The license
• A description of the dataset used to train the model
• Version of the dataset and or tools used to train the model

[robertknight]

Basic support for this was added in #48. I will flesh it out with more fields as needed.