Buffering with a small value is very slow when profiling anything in nvpp.

Amends e5fc6a96

Scan refactoring

Closes #87

See merge request !100

- structs from HorizontalOperations.h reimplemented as function objects in
  Functional.h
- repetitive function definitions generated using macros
- added new operators: % (modulus) and ^ (xor)

This way we test both the general CUDA implementation using shared
memory and the specialization using __shfl instructions.

Both the reduction and scan kernels needed some tweaks due to shared
memory usage with non-fundamental types.

This is needed because custom specializations of std::is_arithmetic
cannot be used (they cause an undefined behaviour).

The algorithms are implemented as plain functions in TNL::Algorithms.
containsValue was replaced with contains.

Removed reduction methods from Array and ArrayView, instead added overloads of reduce and reduceWithArgument for arrays/views

Plain functions are much more flexible than methods. The methods were
also violating the open-closed principle:
https://en.wikipedia.org/wiki/Open%E2%80%93closed_principle

This adds back the original approach (prescan + uniform shift) which
was removed too early.

The original approach (prescan + uniform shift) is more efficient for
inputs that are expensive to evaluate, such as vector expressions.

Using an odd number of valuesPerThread avoids shared memory bank
conflicts even without a special interleaving. We also save some shared
memory this way.

Small inputs can be scanned with just one CUDA block, which avoids the
scan of block results and second-phase kernel. Hence, large arrays can
be scanned with just 3 kernel launches instead of 4.

The input values are first copied into shared memory, reduced
sequentially across chunks, and scanned only at the end of the kernel.
This follows the upsweep-downsweep approach by Blelloch which is more
work-efficient. Also the distinction between exclusive and inclusive
scan appears only at the end of the kernel, which avoids the weird "+2"
size of the shared memory.

Also used Cuda::getInterleaving() for the indices when accessing the
chunkResults array, which avoids shared memory banks conflicts in the
spine-scan phase.

- input and output are passed by views rather than raw pointers (this
  allows to scan even vector expressions)
- consequently, indexing is different (begin and end for the global
  memory accesses)
- fixed calculation of currentSize in the launcher
- changed configuration of the kernel using the blockSize and
  valuesPerThread template parameters rather than the elementsInBlock
  runtime parameter
- changed allocation of the shared memory from dynamic to static
- the second phase kernel uses shared memory to cache block results for
  each block