Use concise names for new public datasets

[dylanuys]

Summary

Rename the 46 newly added public dataset entries to concise, lowercase names. Remove release version and media label prefixes, normalize separators, and shorten the warehouse shard names. For example, v25-synthetic-Persona-Fluxed-10k-2608_brazil-train becomes persona-fluxed-10k-2608-brazil-train.

Only dataset name fields change. Source paths, revisions, filters, labels, and sampling settings remain the same.

Validation

• Confirmed all 46 changes are name-only and the resulting names are unique across the public configs.
• Loaded every public dataset config with GasBench's config loader.
• git diff --check passed.