Disable extracting title from document content by default

See the VOTE and explanations at http://markmail.org/message/lfzwyq6aaftbzhbs

The rationale is:

• This is costing more than it should. It needs parsing the document content to
  get its XDOM and then to traverse it to find Heading blocks. When we display doc
  titles in lists (in livetables for example or in the activity stream) it costs a
  lot to do so.

• Our implementation is broken since if you use an include macro that generates
  a heading for ex, it won’t be taken into account since we don’t apply
  transformations ATM when computing the title since it would be even more
  expensive.

Since 6.x is about performance I believe this would be a good step forward.

Now in order to handle backward compatibility. I propose that we:

• Add a legacy configuration parameter to keep the behavior (but off by
  default). This would be in order to let users using this feature convert their
  wiki
• Move the code to a legacy module (I hope it’s possible)