Pin, exclude, or un-curate one crawled page.

Takes effect on the NEXT crawl, not now: the crawler is the only writer of crawl scope (retrieval/crawler.py reads curation in its cap and its eviction loops), and having the route also delete index rows would give the index two writers with different normalisation. The one exception is the obvious one a customer expects — excluding a page they can see the desk quoting — so that case drops the index row immediately.