Build Interactive Leaderboard

interactive table listing experiment configuration, benchmark and metrics, sortable by benchmark and one representative metric. Top 3 approaches for each benchmark will be listed in the README.md.

Subtask 1: Provide script that updates a dedicated section in README.md with the current best performing approaches that are parsed from a dedicated results folder. Along the lines:
```Python
leaderboard = df.sort_values(by='your_metric', ascending=False)
return leaderboard.to_markdown(index=False)
```

Metric might be 5-Accuracy and top 3 are reported for each benchmark.

Subtask 2: Provide interactive HTML leaderboard linked on README.md that
- is hosted using GitHub Pages.
- is a dynamic web page with JavaScript libraries (DataTables, riversun/sortable-table, etc.) that supports sorting/filtering