chore: remove the 17 retired skills

[nirsha-brd]

What Removes the 17 retired skills: brand-listening, brd-browser-debug, bright-data-best-practices, bright-data-mcp, competitive-intel, data-feeds, design-mirror, discover-api, js-sdk-best-practices, live-research, price-comparison, proxy, python-sdk-best-practices, rag-pipeline, scraper-builder, scraper-studio, seo-audit. Remaining: agent-onboarding, brightdata-cli, scrape, search (their refreshed versions land in the follow-up PRs). The old 21-skill README is swapped for a short interim README, and the .claude-plugin folder is removed entirely: the repo ships plain skills, not a Claude Code plugin. An org-wide code search confirms no repo depends on those manifests, and the CLI never reads them. The one effect is on users who added this repo as a Claude Code plugin marketplace, whose plugin updates stop working. ## Why The retired skills are dead. This is the first PR of an incremental stack that replaces the catalog: next comes agent-onboarding plus the new datasets skill, then the refreshed scrape, search, and brightdata-cli plus the new fetch, browser, billing, brightdata-mcp, and brightdata-sdk skills, and finally the full README and manifests for the complete set. ## CLI impact, verified The CLI reads this repo only during bdata skill add (grep of the whole dist: no other command touches it). Six names in the shipped registry point at removed folders (data-feeds, bright-data-mcp, bright-data-best-practices, design-mirror, python-sdk-best-practices, scraper-builder). After this merge each fails with the CLI's contained per-skill fetch error and a clean summary, no crash, proven by running the real installer against a deleted name. The registry itself is updated by brightdata/cli#27. ## Known interim state The four remaining old skills still contain 65 textual references to removed skills. They are replaced wholesale by the next PRs in the stack, so those references are not patched here.