fix bleurt docs

fix bleurt docs: a task in repo2rlenv-pr-diff (Harbor dataset). Fixes several confusing things about the BLEURT metric - A logging warning states that "BLEURT-Base" is used as the model checkpoint, making it seem like "BLEURT-Base" is a model checkpoint identifier (it's not). 'bleurt-base-128' is…

The task

Fixes several confusing things about the BLEURT metric - A logging warning states that "BLEURT-Base" is used as the model checkpoint, making it seem like "BLEURT-Base" is a model checkpoint identifier (it's not). 'bleurt-base-128' is the actual default - The README mentions a `checkpoint` argument, but the correct…

Part of FineEnvs/repo2rlenv-pr-diff.