THis commit changes the library version to 0.9.0 and the format
 version to 0.6.0. The new format version break compatibility
 which is caused by the new physics.

 There is also an update for the train_properties json
 format in this commit. This just explains the format changes in
 format 0.6.0.

 the unit test failed because the value in the test and in the
 test_train_data_1.json did not match.

Final work of BBB

See merge request !2

changed default value of track_drag, the renaming will come later

added my tested train to core/tests/data in order to get the right
values

moved 2 functions from public to private and removed some lines which
are not needed any more

rewritten the documentation of nearly all functions in physics.hpp
changed the variable named track_drag to the better name of train_drag

Moved some varables from .hpp to void tick() including constant
variables which are defined by Train.json

Added the documentation of the physics.

solved some issues mentioned from @nkrs14

small bugfixes in physics.cpp cause the abs function doesnt work for
speedlevel. The problem of small values is now fixed in a other way

improved the issues @nkrs14 mentioned - smaller bugfixes and removal of
unnecessary lines.

removed some unnecessary, commented lines

when speedlevel is zero nothing will be done in physics to pretend
driving without accelerate

without any key pressed

first implementation of an input_axis and compiler error fix

See merge request lhsw83/libtrainsim!2

before there was an enum which contained a list of all of the possible backends. This is now replaced by a class and some constant objects. This way new backends can easily be defined and other parts of the program can figure out which rendererType and which WindowType is used. An example of this is in the control system. The control system only cares about the window type and can completely ignore the renderer. Because of this no change is needed if a different renderer is used for as long as it is one of the supported windowing systems.
Everything in the video module should now be in the Video namespace except for the video singleton.

Now the rework of the video backend is almost finished. The generic video backend now uses both the generic window manager and generic renderer. This allows a very flexible structure and enables developers to implement their own window manager and by building it on top of these classes.

Next the classes will be moved around in the namespaces and some documentation will be added.

This moves all of the opencv renderer code into a separate class and uses it. This backend might be removed in the future since it is to slow to be really usable but it is required for now.

Now the frame is defined by the renderer it is intended for not by the renderer/window combination. This is another step to enable addition windowing systems like glfw, gtk and others.

now there is a generic window manager and a child for the sdl2 window management. This should make it easy to add new window management APIs. Once the opencv backend is ported over, there old class will be removed.

there is now a generic renderer and all of the ffmpeg code was moved into a child of that generic renderer.
The ffmpeg_sdl backend is now very small thanks to those changes. The next thing to come is a generic window management and adapting the opencv backend to this new structure. This might allow opencv/ffmpeg and more alternatives to sdl2