moved generic blockforest creation functions from pe/CreateWorld to blockforest/Initialization.h

use the general classes instead. they support levelwise load
balancing via a constructor parameter

default is levelwise for backward compatibility

- information from CMake to codegen: double/float, OpenMP, ...

- intel 19 internal bug when using auto, and implicit capture in lambdas

- LB on GPU, Uniform Grid Benchmark app
- helper class to schedule tasks to multiple CUDA streams

- 20 template parameter ought to be enough for everybody :P