Fine-tuning with OpenAI

[seancsmith]

• Work out how to fine-tune the model on the 10 examples used for annotation to compare against few-shot method.
• If improvement is shown (or even if it's super cheap) train on all 45 annotated studies

[seancsmith]

• created a workflow to create training data in correct format and build fine-tuning jobs. Slight improvement was shown using the 10 initial annotations as examples. Will be interesting to see results using a much larger training set. Will create a new issue for training on larger sample.