websearch: detect anti-bot/captcha pages and surface clear error

DuckDuckGo (and Bing) serve anti-bot challenge pages with HTTP 200 when
a request is flagged (e.g. TLS fingerprinting / IP reputation on Linux,
GH #270). Previously these parsed to an empty result set that was
indistinguishable from a genuine zero-result query, so the failure was
silent and the user just saw 'No results found'.

Add detect_anti_bot_page() which recognizes common challenge markers
(DDG anomaly modal, captcha/recaptcha, cloudflare challenge, human
verification). When DDG/Bing return zero results AND a challenge page is
detected, return a descriptive error instead. This both surfaces the
real cause and triggers the existing fallback engine chain.