docs: rewrite plugin docs to match current app behavior (#2506)

Co-authored-by: Claude <noreply@anthropic.com>

docs/website-tutorial.md

@@ -18,20 +18,28 @@ A comprehensive guide to testing your LNReader plugins using the web interface.
 
 ## Features Overview
 
-The testing website provides five main sections to test different plugin functions:
+The testing website provides five tabs to test different plugin functions:
 
-- **Headers** - Configure custom HTTP headers
-- **Popular Novels** - Test `popularNovels()` with pagination and filters
-- **Search Novels** - Test `searchNovels()` with search queries
-- **Parse Novel** - Test `parseNovel()` with a novel path
+- **Popular** - Test `popularNovels()`, including its `Latest`/`Popular` toggle, page-by-page
+  fetching, and filters (via the `Filters` button, when the plugin declares any)
+- **Search** - Test `searchNovels()` with search queries
+- **Parse Novel** - Test `parseNovel()` with a novel path. If the plugin sets `totalPages` (see
+  [Pagination](./docs.md#pagination)), this tab also exercises `parsePage()` through
+  Previous/Next/Fetch Page controls, and offers an "Export EPUB" button that fetches every
+  page's chapters (via `parsePage`, when paginated) and each one's content (via `parseChapter`)
+  into a downloadable EPUB file
 - **Parse Chapter** - Test `parseChapter()` with a chapter path
+- **Settings** - Playground-wide request configuration: the browser User-Agent (and whether to
+  send it), extra cookies to attach to every request, and the fetch mode (Proxy/Node
+  Fetch/Curl) used to reach the target site. This is a testing convenience for the playground
+  itself, separate from a plugin's own [`pluginSettings`](./docs.md#pluginsettings)
 
 ## Pre-Submission Testing
 
 Before submitting your plugin, verify that:
 
-- All five sections work without errors
-- Multiple pages load correctly
+- All five tabs work without errors
+- Multiple pages load correctly, for both `popularNovels` and, if implemented, `parsePage`
 - Search returns accurate results
 - Novel parsing extracts all metadata
 - Chapter content is clean