Refactor Mentionable

Caching is now always enabled when rendering attributes as there's no
reason to not cache them. The method Mentionable#all_references has also
been modified to take an optional Gitlab::ReferenceExtractor instance to
allow other code to create one and re-use it between multiple calls.

app/models/concerns/mentionable.rb

@@ -23,7 +23,7 @@ module Mentionable
 
   included do
     if self < Participable
-      participant ->(current_user) { mentioned_users(current_user) }
+      participant -> (user, ext) { all_references(user, extractor: ext) }
     end
   end
 
@@ -43,23 +43,22 @@ module Mentionable
     self
   end
 
-  def all_references(current_user = nil, text = nil)
-    ext = Gitlab::ReferenceExtractor.new(self.project, current_user || self.author, self.author)
+  def all_references(current_user = nil, text = nil, extractor: nil)
+    extractor ||= Gitlab::ReferenceExtractor.
+      new(project, current_user || author)
 
     if text
-      ext.analyze(text)
+      extractor.analyze(text, author: author)
     else
       self.class.mentionable_attrs.each do |attr, options|
-        text = send(attr)
+        text = __send__(attr)
+        options = options.merge(cache_key: [self, attr], author: author)
 
-        context = options.dup
-        context[:cache_key] = [self, attr] if context.delete(:cache) && self.persisted?
-
-        ext.analyze(text, context)
+        extractor.analyze(text, options)
       end
     end
 
-    ext
+    extractor
   end
 
   def mentioned_users(current_user = nil)