Include human curation vs LLM curation for prognostic evidence analysis

[korikuzma]

From @ahwagner :

We still need to include the 75 manually-reviewed records used in the AI model evaluation for prognostic variants. We are clear to include these in our GH repo, as well as the analysis notebooks used for the results.