FIX: Unescapes hash section with present to account for url-encoded chars

Sections with unreserverd characters will appear url-encoded and need to
be unescaped before using it.

Wikipedia generates 2 different spans in this case in the same page, one
with an id resulting of replacing the % symbols with . and the other with
the decoded version of the string. For example, for /wiki/foo#A%C3%A1A it
will generate:

<span id="A.C3.A1A"></span>
<span id="AáA">AáA</span>

Unescaping the `m_url_hash_name` should work in all cases to target the
proper section span.

lib/onebox/engine/wikipedia_onebox.rb

@@ -24,7 +24,7 @@ module Onebox
         end
 
         unless m_url_hash.nil?
-          section_header_title = raw.xpath("//span[@id='#{m_url_hash_name}']")
+          section_header_title = raw.xpath("//span[@id='#{CGI.unescape(m_url_hash_name)}']")
 
           if section_header_title.empty?
             paras = raw.search("p") # default get all the paras