This tests are designed to guarantee that the wide lines behave exactly
like normal lines drawn with the Bresenham's algorithm.
This tests are somewhat subjective since this is non-defined behavior,
but I think that mimic the Bresenham's algorithm is reliable enough.
Currently the horizontal version of this test **fail**.
Only the oblique 3 pixels wide lines are defined:
* The oblique 2 pixels wide lines are somewhat hard to define.
* To define the oblique lines wider than 3 pixels we neet to define
first how the oblique lines should expand their width (realistic or
exact).
Notice that the expansion of the line width depends on the order of the
points:
* If the bigger axis value is provided as the *second* point the line
expand first to the *positive* side of the axis.
* If the bigger axis value is provided as the *first* point the line
expand first to the *negative* side of the axis.
* If the line's width is odd this doesn't matter, as the line will
expand the same amount to both sides.
This behavior should be consistent in both horizontal and vertical lines.
This is Christoph Gohlke's test suite from his personal PIL package found
at http://www.lfd.uci.edu/~gohlke/pythonlibs/.
This is just to bring it in as a separate commit. Future commits will align
it with Pillow.
