Rename crawled_swebench source to crawled in all dataset splits

Add CRJudgeBench paper citation to dataset card

Remove limitations section from dataset card

Remove PR and repository count columns from split table

Use trustworthy terminology consistently in dataset card

Remove construction and repository coverage sections from dataset card

Describe the task as judging technical trustworthiness

Add dataset card with verified statistics and evaluation guidance

Upload train, validation, and test datasets

initial commit