Rework JS static site code structure

This splits up the generator class as was a problem suspected in this PR
review:

#15640 (review)

It creates a single class for a Browser which extends the
capybara-webkit browser to include a way to get a page's contents after
jQuery's AJAX calls are done.

This is in a StaticSite module just to namespace it and indicate where
it's used in the application, rather than it's nature. In other words
it's not a _static site browser_ but you could read it that way so maybe
this needs changing?

All the remaining procedural logic is pushed into the Rake task. From
here we could then extract that if we see fit.

Rakefile

@@ -57,12 +57,21 @@ end
 require 'bundler/audit/task'
 Bundler::Audit::Task.new
 
-require_relative 'lib/static_site_generator.rb'
+require_relative 'lib/static_site/browser.rb'
+require 'uri'
+require 'pathname'
 desc 'Build static site pages that rely on JavaScript'
 task :generate_static_site_javascript_pages, [:base_url] do |_, args|
   base_url = args.fetch(:base_url, 'http://localhost:4567')
-  javascript_pages_to_scrape = ["#{base_url}/needed.html"]
-  StaticSiteGenerator.new(urls: javascript_pages_to_scrape).build
+  pages_to_scrape = ["#{base_url}/needed.html"].map { |u| URI.parse(u) }
+  browser = StaticSite::Browser.new
+
+  pages_to_scrape.each do |url|
+    browser.visit(url)
+    file = Pathname.new(url.path[1..-1]).sub_ext('.html')
+    file.dirname.mkpath
+    file.write(browser.body_after_jquery_ajax)
+  end
 end
 
 task default: ['test', 'rubocop', 'bundle:audit']

lib/static_site/browser.rb

@@ -0,0 +1,31 @@
+require 'capybara-webkit'
+
+module StaticSite
+  class Browser < Capybara::Webkit::Browser
+    JQUERY_WAIT_TIME = 2
+
+    def initialize
+      super(Capybara::Webkit::Connection.new(server: Capybara::Webkit::Server.new))
+      allow_unknown_urls
+    end
+
+    def body_after_jquery_ajax
+      wait_for_jquery_ajax
+      restore_pre_js_page_classes
+      body
+    end
+
+    def wait_for_jquery_ajax
+      Timeout.timeout(JQUERY_WAIT_TIME) do
+        # Errors if jQuery isn't on the page - should we check for that?
+        loop until evaluate_script('jQuery.active').zero?
+      end
+    end
+
+    # Restores page classes modified by running JS on the page
+    def restore_pre_js_page_classes
+      execute_script("$('html').addClass('no-js')")
+      execute_script("$('html').removeClass('flexwrap')")
+    end
+  end
+end