- 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

- comparison to old CPU transfer communication included

- CUDA compiler has problems to compile external json library
- timing pool is included by BlockForest, and previously timing pool
  included JSON library

previously the ghost layer was aligned

Features:
   - uses generated pack infos for packing & unpacking directly on GPU
   - can directly send GPU buffers if cuda-enabled MPI is available,
     otherwise the packed buffers are transfered to CPU first
   - communication hiding with cuda streams: communication can be run
     asynchronously - especially useful when compute kernel is also
     split up into inner and outer part

- added RAII classes for CUDA streams and events
- equivalence test that checks if generated CPU and GPU (overlapped)
  versions are computing same result as normal waLBerla LBM kernel

- templated buffer classes allow the buffer systems to work with custom
  buffers, e.g. buffers of GPU memory

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

add pretest stage to reduce jobs in pipeline

See merge request walberla/walberla!144

Resolve "Bug in getFriction function"

Closes #72

See merge request walberla/walberla!143

CI: GUI and CUDA

See merge request walberla/walberla!141

-> intel compiler went out of memory

- recent GCC version produce error when "-isystem /usr/include" is added
  to command line. Error is: '#include_next <math.h>' not found
- the removed "hack" added the cuda include dir with "-isystem" to
  the CUDA compiler arguments to prevent warnings from boost header
  -> currently there seem to be no warnings generated from boost headers
     so "hack" is removed

Add Intel 2019 compiler to CI

See merge request walberla/walberla!138