Moved tnlObject_impl.h back to tnlObject.cpp since function definitions
in header files cause duplicates when building libtnl. The constructor
of tnlObject is __cuda_callable__ so it's empty definition was moved
directly to tnlObject.h to make it visible to nvcc.

Adding explicit template instantiation.
Moving some _impl.h files into implementation dir.

Renaming tnlArray to tnlMultiArray.

Finishing new template interface for GMRES.

Adding new class tnlArrayManager for managing arrays - allocation/deallocation, load and save functions etc.
tnlVector is derived from tnlArrayManager.