Improve eval-results badge docs and add moderation note

[gary149]

Summary

• Clarify badge table descriptions to match actual implementation in moon-landing (e.g. verified = valid verifyToken reproduced via Inspect AI, community = open PR, leaderboard = benchmark has a leaderboard, source = external URL provided)
• Add a tip after Community Contributions explaining how community scores can be moderated (author can close the PR to remove a disputed score)

Context

User feedback surfaced two common questions:

1. "Who runs the evals?" — badge descriptions now make provenance clearer
2. "Can someone submit false scores?" — new tip explains the PR-based moderation model

Test plan

• Review rendered markdown for clarity

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.