New Core API for recursive processing

By nature, inside Nuxeo Platform we have a lot of processing that is done on a document sub tree :

• update ACLs on a sub tree (and compute readacls)
• change life cycle on a sub tree
• remove documents (delete a sub tree)
• move a sub tree (and recompute all sub paths)

All these processing have common characteristics :

• they can be long running process
• they can be configured / plugged (ex: version policy removal, lifecycle transitions ...)
• we must ensure they will run at some point

Currently, the majority of these processing are handled at database level using triggers and stored procedure.
This approach shows several limitations :

• since most processing are synchronous it creates freezes for the users
  • a lots of support client have complained about that
• it creates performance issues, especially on database engine that don't handle correctly concurrency
  • we had a lot of problems with SQL Server, but there are also problems with PGSQL
• we need to code, debug and maintain for several database vendors
  • that's really a pain
• because processing is handled at SQL level
  • some configuration rules are by-passed
  • listeners don't get fired (ex: document delete event )
  • we can not use the same approach for non SQL DB anyway !

To fix that we need extract a new set of core level APIs (resurseXXX) that :

• will be shared across most repository backends
  • but may actually be optimized at low level if needed
• will be Java based and will leverage the `WorkManager` infrastructure

Part of the work could also be shared with what has been done for ElasticSearch (recursive re-indexing).