Modify inter-annotator agreement computation

[kasnerz]

Before

Until now, factgenie provided a way to compute Pearson r correlation on dataset- and example-level. The results were displayed in a pop-up with some basic comments.

❌ This was not great for multiple reasons.

First, it impovered the users: it gave them less control over the computed results (unless they wanted to look into and/or modify factgenie code).

Second, it turned out to be inappropriate for long-running computations (such as computing the γ score), which would require displaying a separate page with asynchronous updates.

After

We decided to keep it a simple. On the IAA page, we only provide the users with CSV data files on which they can run data analysis themselves 🔎

We also provide a 📕 Jupyter notebook using the above files. The notebook goes through all the nooks and crannies of IAA computation with detailed comments.

This pull request resolves #99.