Why do problems have this method anyway? It was obviously never used...

When type traits for the "double-float" or "double-double" types are
needed, they should be added as specializations of the
std::numeric_limits template in some of the TNL header files.

There were 2 problems:

- it does not belong into a numerical library
- the approach with hardcoded escape codes is not portable even on Linux

There are separate libraries which solve both of these issues much more
nicely, they should be used instead.

It is not used in current code, it was used only in some legacy benchmarks.

* origin/mesh: (251 commits)
  Updated CMakeLists.txt files for meshes
  Added missing __cuda_callable__ flag to MeshEntity
  Fixed comparison between signed and unsigned
  Optimizing mesh initializer, replaced EntitySeedKey with EntitySeedHash
  Added timer to MeshReaderTest
  Fixed SubentityAccessLayer
  Fixed typo in mesh StorageLayer
  Removed checkCells method from the mesh initializer
  Switched static functions to inline
  Fixed comparison between signed and unsigned
  Fixed constness of reader in MeshTypeResolver
  Renamed VTKEntityType to EntityShape
  Renamed MeshConfigBase to DefaultConfig
  Removed static asserts from MeshConfigBase
  Fixed internal linkage of functions in VTKEntityType.h
  Added specialization of getEntityMeasure for quadrilaterals
  Simplified and generalized the getEntityCenter function
  Added getEntityCenter and getEntityMeasure functions
  Removed unused methods
  Removed GlobalIndexType and LocalIndexType from the public interface of MeshEntityTraits
  ...

* master:
  Documentation.

Python bindings

See merge request mmg/tnl-dev!2

And of course our implementation is faster ;-)

It cannot work because indices are not initialized to -1 anymore. The
checks in the MeshBuilder class should be enough.

This way GCC does not complain that they are not used in some
compilation units.

It should not be called "base" because inheritance does not work for the
constexpr methods.