Set up website-crawl knowledge base

Configure the website-crawl knowledge base for an agent. PUT semantics: the request fully replaces the set of crawl roots. For each rootUrl already present in the agent's configuration, the existing dashboard-set per-page selections (nodeSelections) are preserved and only excludedUrls are overwritten. A rootUrl with no per-page selections is seeded with its root page, which the crawl starts from and expands to discover linked pages beneath it. Existing rootUrls absent from the request are removed. Pass an empty configs array to stop crawling entirely. A website sync is enqueued on every successful PUT.