Make model_config more consistent

[MattFisher]

Need a review on this one before regenerating the dashboard logs

[eleweek]

LGTM overall, but requires fixing the tests

[celiawaggoner]

nit: remove https://github.com/ArcadiaImpact/inspect_evals_scoring/pull/19/files#diff-d77ad65e926a80f1c601cea082ce57e74896a5d447a07b845d73dddf123d22faR208-R223 or update it so that whenever we do un-comment it, it's correct

looks good to me, with the one suggested above about the commented code

[MattFisher]

Okay so the .dashboard.json files are now all updates in the stage folder, so should be showing up on the staging dashboard at https://inspect-evals-dashboard-stage.streamlit.app/ , but the values don't seem to have updated yet so I think they're still cached.

Once we've QA'd the staging site we can move the updated logs across to the prod folder in the S3 bucket.

(A good check is that the "API Provider" for Anthropic models has changed from "Anthropic (via AWS Bedrock & Anthropic API)" to just "Anthropic")