Fixed handling of the boundary conditions for explicit solver - introduced a method Problem::applyBoundaryConditions.

Tunning experiments - 'pure-c' benchmark copied to 'tunning' and refactored to work with operator and boundary conditions.

Implemented CUDA SharedPointer with CUDA Unified Memory. Does not work with sm30 and it is very slow.

* cmake:
  Added --offline-build option to skip update of external Cmake projects
  Disabled linking to the VTK library and enabled all mesh tests for CUDA even for nvcc
  Removed check for HAVE_CUSPARSE
  Declared all custom build options in CMakeLists.txt

* common-data:
  Re-enabled Python build in CMakeLists.txt
  Refactoring done - MeshDependentData was removed.
  Removing mesh dependent data.

It is not needed, since CUSPARSE is included in all versions of CUDA
since at least 7.0.

Also note that CUSPARSE is needed only for some tests and legacy code
(tnl-benchmark-spmv and wrappers in Matrices::CSR) and experimental code
(Preconditioners::ILU0). Everything is still guarded by the
HAVE_CUSPARSE macro, so the user can pass -DHAVE_CUSPARSE -lcusparse to
the compiler to enable the features.

This way they can use either yes/no, on/off, or true/false.