Add arg to force comparison and always register/deploy model

[jfomhover]

PR into Azure/mlops-v2

here's a summary of the issue:

• model deployment in the training pipeline is conditioned to a test comparing the perf of the fine tuned model with the non-fine-tuned model
• in theory the fine-tuned model should win, but we're using a small SKU to allow everyone to run this NLP scenario without specific quota, and we limited the number of data to 1000 and epochs to 5 so that the training pipeline would not take hours
• when running under those conditions, it is likely both models will perform pretty much the same, making comparison between model uncertain
• also the baseline (non-fine-tuned) model is already pretty good in itself, making this comparison uncertain as well

We're proposing to force the comparison by default to switch to an always-deploy scenario, but leave the pipeline as it is so that users can adapt to their needs more easily and see the full extent of the pipeline.

Checklist

I have:

• read and followed the contributing guidelines

Changes

• add argument in comparison script to force comparison and force register the model