feat: add scrapeunblocker plugin

[Kontuzijus]

Plugin Information

• Author: scrapeunblocker
• Plugin name: scrapeunblocker
• Version: 0.0.1
• Source repository: source is included in this PR under scrapeunblocker/scrapeunblocker/
• Contact: https://www.scrapeunblocker.com

Submission type

• New plugin

Description

ScrapeUnblocker renders web pages in a real browser behind anti-bot protections
(Cloudflare, DataDome, PerimeterX, Akamai) and returns the result to a Dify
workflow. It is useful when an ordinary HTTP request returns a block page, a
captcha, or an empty shell instead of content.

Two tools:

• Get Page Source - fetch a single page. Returns rendered HTML, or AI-parsed
  structured JSON when parsed_data is set. Supports proxy_country for
  geo-restricted content.
• Search Google - scrape a Google SERP for a keyword and return the organic
  results (title, url, description, position).

Risk level

• High risk

This plugin proxies and fetches arbitrary user-supplied URLs, so it falls under
the high-risk category by your review guidelines. Specifically:

• Arbitrary URL fetching is the plugin's purpose. Requests are made by
  ScrapeUnblocker's infrastructure, not by the Dify host, so the target site
  sees a ScrapeUnblocker IP.
• Timeouts are explicit on every request (requests.post(..., timeout=...),
  default 180s, configurable via the timeout_seconds credential). Nothing can
  hang indefinitely.
• The API key is never returned in errors. Network failures raise only the
  exception type and the path; HTTP failures raise only the status code and
  path. Request headers, which carry the key, are never interpolated into
  messages.
• Data flow is documented in PRIVACY.md: the target URL (including its
  query string), the optional country code, the search keyword, and the API key
  are what leave the Dify instance; the fetched page content passes through
  ScrapeUnblocker's servers. It also warns against putting secrets in target
  URLs.

Verification

The client was exercised against the live ScrapeUnblocker API before packaging:
get_page_source returned a fully rendered page (58,917 chars, correct title),
search_google returned 20 organic results with the documented fields, and
validate_credentials succeeded. An invalid key correctly raised a credential
error, and the test asserted the key does not appear in the error text.

Provider and tool classes were imported against dify_plugin 0.9.x to confirm
the SDK surface used here exists. Package built with dify plugin package
(CLI v0.6.4); the archive contains 14 files and no build artefacts.