Create baseline gneissweb science score of selected top10k list

[dalhuijsen]

Generate mean/median/maximum gneissweb science scores of base crawl and focus crawl and compare differences to get an indication of success.

[malteos]

This PR contains a first evaluation: #20

Comparison: CC-SUPPLEMENTAL-2026-22 vs CC-MAIN-2024-18  [real focus-crawl data]
  Direction:     focus is higher than baseline
  Mean delta:    +0.403044
  Median delta:  +0.666728

See https://github.com/commoncrawl/crawl-openathena/blob/feat/classify-warc/notebooks/compare_classifier_scores.ipynb

[malteos]

Notebooks are merged to main: https://github.com/commoncrawl/crawl-openathena/tree/main/notebooks