* parallel-reduction:
  Fixed ArrayOperations::compareMemory for MIC
  Fixed parallel reduction for MIC
  General parallel reduction on host
  Refactoring parallel reduction operations

- used variadic templates to avoid code bloat
- added execHost static method to avoid nvcc warnings

- separate data types for input arrays of binary operations
  (necessary for the equality comparison of two arrays of different
  type)
- added separate template parameter for ResultType, which defaults to
  the RealType
  (the goal is to allow sum of bool vectors into int instead of bool)

Long live clang for pointing this out!

Standard CMAKE_CXX_FLAGS are a space-delimited list, but our
CXX_TESTS_FLAGS are manually passed to CUDA_ADD_EXECUTABLE, which
expects a semicolon-delimited list. Blame cmake for such a crazy
language...

- Use math function from TNL/Math.h instead of the standard library
- Many cosmetic changes for readability

- The inherited operator= is weird, but this seems to work.
- Use TNL::pow instead of std::pow etc.

The splitting with if( std::is_same<...> ) is not perfect, the code for
all devices was always visible to the compiler, and irrelevant parts
were optimized out at a much later stage. To fix compilation for MIC and
custom data type, we need proper template specializations.

- Cleaned up unnecessary specializations of TNL::max and TNL::min
- Implemented result type promotion to fix compilation with nvcc
- Added TNL::sqrt, which will be necessary for MIC

It used to work everywhere, but the compilers on CentOS (on penny)
are weird...

Conflicts:
	README