Add corrected v3 OpenEval outputs

[sbooeshaghi]

Summary

• add OpenEval outputs for the latest genuine versions of elife-60151 and elife-78908
• preserve all existing v2 outputs unchanged
• include claim extraction, OpenEval evaluation, positional claims, and run metrics for each v3 manuscript

Selection rule

The PRC benchmark uses the earliest genuine manuscript version. The OpenEval-only corpus uses the latest genuine manuscript version. Auditing the frozen cohort identified these two manuscripts as the only selected-version changes required by that rule.

Validation

• 60151 v3: 108 claims and 21 OpenEval results
• 78908 v3: 64 claims and 11 OpenEval results
• output and prompt hashes are recorded in the reproducibility repository
• v2 artifacts were not modified or removed