The dataset card points readers at open-markdown-v2, but export publishes v3 #88

Problem

markdown export --repo defaults to open-index/open-markdown-v3 and the schema is open-markdown-v3, but GenerateMarkdownREADME hardcodes open-index/open-markdown-v2 in eight places: the datasets snippet, the huggingface_hub snippet, the local_dir, the DuckDB path, the homepage link, and the discussions link.

So the card committed into the v3 repo tells a reader to download v2. Every download snippet on the page is wrong for the dataset it is on.

Found while adding the extractor column in #49. Left out of that PR to keep it to one thing.

What to do

Add Repo to MarkdownDatasetStats and thread cfg.Repo through the committer, then use it in the snippets instead of the literal. A run publishing to somebody's own repo should get snippets that point at their repo, not at ours, which is the reason to plumb it rather than change the literal to v3.

Done when

• a card generated for a run with --repo org/name names org/name in every snippet and link
• a card generated with no repo set falls back to the export default