fix: classify media and social catalog entries

[kriptoburak]

Summary

• keep broad media keywords in the multimedia category instead of the entertainment category
• add social platform keywords for Twitter/X, tweets, Weibo, and WeChat catalog entries
• export the merge classifier and add regression coverage for an X/Twitter data tool, a generic media player, and a Steam game helper
• make the MD8 design hook tolerate clean clones where outputs/reports has not been created yet

Validation

• node --test tests/merge-all-sources.test.mjs
• node --check scripts/merge-all-sources.mjs
• node --check scripts/md8-design-hook.mjs
• npm run build
• git diff --check
• changed-surface URL check: reachable URLs returned 200; placeholder clone URL and unresolved Pages hostname were treated as existing docs state, not changed content