Restructure for multi-config: default/ and claude/ subdirectories

Fix multi-config: remove loading script, add dataset_infos.json

Add dataset loading script for multi-config support

Simplify dataset metadata for Hub viewer compatibility

Add Claude-level dataset: 30k records with deep reasoning

Fix README metadata: use configs for JSONL dataset compatibility

Initial release: 100k Japanese coding excellence dataset