Moved (most of) static methods from TNL::Devices::Cuda as free functions into separate namespace TNL::Cuda

The class TNL::Devices::Cuda was too bloated, breaking the Single
Responsibility Principle. It should be used only for template
specializations and other things common to all devices.

The functions in MemoryHelpers.h are deprecated, smart pointers should
be used instead.

The functions in LaunchHelpers.h are temporary, more refactoring is
needed with respect to execution policies and custom launch parameters.

Fixes #46

The x and b vectors might have different type.

- base class with non-virtual destructor causes undefined behaviour
- missing return statements at the end of non-void functions
- missing const before char* function argument
- reordered initialization of class members
- misleading indentation after for statement
- fixed use of uninitialized variables
- fixed unsequenced modification and access to variables (the C++
  standard does not define the evaluation order of operands, so
  expressions with side-effects such as j++ cause undefined behaviour)

Implemeting heat equation benchmark.

RowLengths was renamed to CompressedRowsLengths.
OpenMP support wads added to matrix-vector multiplication and to scalar
products.
CSR format can use Cusparse for SpMV.

Min. interations feature was fixed in the GMRES method.
Timers for explicit and semi-implicit timestepping were added.

Adding --fast-build optin for the build script to speed-up compilation.