This prevents a bug where a simplified stroke (containing only one
polygon) was incorrectly saved as a polyline, which meant the stroke was
lost when the document was loaded.

Up until now, the fill rule of a polygon was always saved as even-odd,
despite the fact that in most if not all cases, polygons are drawn with
winding fill within OpenBoard.
Saving is now fixed, but there is no way to know upon loading whether
the polygon was correctly saved or whether; so for now, we just set the
fill rule to Winding when loading a polygon.

If enabled in the preferences menu, pen and marker strokes will be
replaced by a simplified stroke after they are drawn.

The algorithm is very basic (for now): if three points are almost lined
up (the threshold angle can be specified in the config file), then the
middle one is removed. This is repeated over the whole stroke; new
polygons are then generated based on the simplified stroke points.

This typically cuts down on number of points and polygons by a factor of
about 10, while having minimal visual impact.

Due to antialiasing, adjacent polygons are separated by a very fine
space. The previous solution attempted to hide this by adding a border
to the polygons. The border of adjacent polygons would overlap, which
was visible (despite the attempted color correction) and, more
importantly, caused massive lags especially on Linux.

Therefore it has been removed but feel free to revert this commit some
day and try to fix this more cleanly.

Several issues remain with multi-screen mode on Linux. The behavior is
inconsistent from one desktop evironment to the next, making it hard to
work around these problems. Among the known issues at this stage:

On Ubuntu 14.04, a call to QWidget::setGeometry requires the widget to
be hidden on KDE, but visible on MATE, for the geometry changes to take
effect.
Despite the widget's geometry being updated by this call, the windows
aren't necessarily moved. Meaning that the control and display widgets
will tend to be displayed on the same monitor, even though their
positions are correctly set to different areas on the extended screen.

In the current state, this behavior is observed on MATE. Unity works
fine and KDE only has transient positioning issues (for example,
swapping control and display windows in multi-screen mode leads to both
windows being placed on the same monitor, until multi-screen mode is
turned off then on again).

# Please enter the commit message for your changes. Lines starting
# with '#' will be ignored, and an empty message aborts the commit.
# On branch dev
# Your branch is ahead of 'origin/dev' by 29 commits.
#   (use "git push" to publish your local commits)
#
# Changes to be committed:
#	modified:   src/core/UBApplicationController.cpp
#	modified:   src/core/UBDisplayManager.cpp
#	modified:   src/core/UBDisplayManager.h
#
# Changes not staged for commit:
#	modified:   release_scripts/linux/build.sh
#	modified:   release_scripts/linux/package.sh
#
# Untracked files:
#	release_scripts/linux/generateDependencies.sh
#

text items being editable on page load

In the right-hand pane, two folders that were at the same path and whose
names started with the same characters were considered to be nested by
the breadcrumbs trail.

E.g, folders named "abc" and "abcd", both in the "Audio" folder:
clicking on "abcd" made the breadcrumb trail display "[Audio] > [abc] >
[abcd]"

On OS X, making annotations in desktop mode then switching to board
mode and back could cause shadows to be drawn around every stroke, and
these persisted after erasing the strokes (though they did disappear
upon switching to board mode and back again).

Fixed type testing of media items (qgraphicsitem_cast couldn't tell the difference between audio and video items)

Adds (experimental) stroke simplification methods