reference performance of AI models on coding tasks

we copy-paste notable coding leaderboard snapshots, with source

this will be useful to argue for the choice of models in our future papers.

The issue names no file, test, or entry point. First locate where benchmark or model references are maintained, then review the existing format before adding dated coding-task leaderboard snapshots with their sources. Done means the selected snapshots are recorded consistently and can support future model-choice discussions.

Written by the indexing model from the issue text.