ZP-652 #comment better diffstate performance

* commit '8cd0a71f':
  - improve performance, by not iterating for each entry through all entries, but initially create an associative array of old entries indexed by their key and then iterating through the new entries once identifying new, modfied or removed entries (this makes a hugh difference eg. for addressbooks with a couple of thousand entries more then a factor of 10) - it drops the requirement that ids need to be in ascending order, new diff code works also if $new contains smaller ids. That happens eg. for us when a users unsubscribs from a mail folder and subscribes again

(cherry picked from commit 8f0ed00c)