Rendered tags

Major refactoring of the interface and the implementation

The API is now in one file, declaring two classes and one type alias. The
classes provide respectively the detection of chilitags in an image, and the
estimation of their 3D pose. The type alias represent the coordinates of the
four corners of (2D) tags. The classes are default contructible, with
reasonnable default parameters, which can be set up by calling methods (e.g.
filtering, camera calibration...).

There is no more "ChilitagsRegistrar" singleton, the result of the
detection/pose estimation is simply returned as a map from the corresponding
method.

The use of each class is demonstrated in two samples each: a simple, CLI-based
one, and a graphical one.

Internally, functors structure the detection, instead of "Pipables"

The naming convention also changed
from (pFunctionParameter, tLocalVariable, mClassMember, scConstant)
  to ( functionParameter,  localVariable, mClassMember,   CONSTANT)
and indentation is now done with 4 spaces.

There are more automated tests too, and bug fixes.