Reimplemented Config::ConfigDescription and Config::ParameterContainer using std::vector instead of TNL::List

This is much safer and simpler because it solves many problems due to
cyclic inclusion of TNL headers.

TNL is not supposed to have any external dependencies, so an empty
config file is useless. If some marginal parts of the project (such as
code for image processing) need to depend on some non-standard library,
they should be split into a separate module/project so that the
dependencies can be properly specified.

For now, the macros are specified as compiler parameters. Client
software using the affected headers will have to do the same.

They are part of POSIX so they should work on anything but Windows. This
can be checked simply by a predefined macro, configuration during build
is not necessary. See https://stackoverflow.com/a/4575466 for reference.

fixes #13

The former has been formally deprecated for a long time and in CUDA 10.0
it is finally properly marked as deprecated, so the compiler can print
warnings.

- There is an RAII wrapper called ScopedInitializer to handle
  finalization due to C++ exceptions and to avoid calls to Finalize in
  all exit branches.
- Fixed signature of MpiCommunicator::Init to match that of MPI_Init.

The std::isnan method can be safely used in host code (if the compiler
is not moron). There is no device code which uses isnan so the
TNL::isNan function is useless.

This reverts commit 0eff2f63.

It has not been implemented in the past decade so it can be safely
assumed that it is useless.