- new representation of task that allows to decrease overhead in task creation
- suppress static graph (computation of successors). It will be added after
without copying the graph representation.
- suppress CUDA support. It will be added after as a pluging
- add support for parallel region in sens of OpenMP
- add new implementation of for_each with several scheduling strategies:
	* Static' OpenMP compliant implementation
	* Steal : work stealing based
	* Adaptive : see IWOMP2013
- compact code and suppress deadcode