Fixes #42

Benchmarks can be easily profiled even without this parameter, so it was
just an unnecessary complication.

This is nicer because it more clearly separates data load, computation
and data store. Furthermore, it allows to use instances of std::plus,
std::logical_and, std::logical_or, etc. instead of custom lambda
functions.

It contained only methods for prefixSum and segmentedPrefixSum, which
were identical for Host and Cuda, so they can be easily implemented
directly in Vector and VectorView.

The performance seems to be identical to the code using volatile.

They are not implemented anyway...

This is necessary to be able to run tests in parallel.

The build takes too long because of this and nobody uses it anyway. CUDA
does not support long double in device code at all.

NDArray

See merge request !18

DistributedNDArray: added methods getLocalIndexer, getLocalView, getConstLocalView, getStorageIndex and getData

This is necessary for local indexing of DistributedNDArray.

It works for any value, device and index types, but the permutations of
both arrays must be the same and both arrays have to be contiguous.