New features
* Extend marker system: Allow unlimited markers and zoomviews
* Extend math channels: Dynamic math channels. Math Modes: +,-,*
* Add a self-calibration for gnd-offset/gain-limit values.

Bug fix
* Export: Always keep a pointer to the the last sampleSet. This ways exports work if the user uses a one-time trigger and want to export the visible graphs.
* Disable roll-mode for DSO6022

Documentation
* API of DsoControl documented and slightly changed.

Easier maintenance
* DsoControl: Samples are always in [-1,1] range now and user-gain/offsets are applied in post/graphgenerator.
* Qt3d is used for the scope window now. No direct shader programming for different platforms or GL versions anymore.
* Move settings related files to own subfolder
* Make GUI docks independant objects. They react on settings changes and manipulate settings or the dsocontrol object directly.
* Split HantekDsoControl into DsoControl+DsoLoop+DsoCommandQueue
* sispinbox does not allow a fixed set of values anymore. Made it over complicated and we have QComboBox for this purpose anyway.

Optimize
* Allow DSO models to not always have 9 fixed gain steps. The DSO6022 for example only has 2.
* Remove a lot of allocations in the fetch-samples/convert/show-graph hot-path. We have a PostProcessingResult pool now.
* Use the new GlScope class to draw the export graph, no dublicate code and it does look exactly like on screen (except different dimensions).

…s well

 * rename post/enums.h to enums-post.h so that MOC won't overwrite
   the file produced for hantekdso/enums.h
 * explicitely construct unique_ptr when insering sliders

Make ChannelWidgets look more like a composite widget and layout
properly. Previously it was a widget which was never added to any layout.
Therefore it had default position (0,0) and size (100, 30) inside
it's parrent (DsoWidget). Whenever a trigger/level slider was in
the left top corner, it was getting under invisible ChannelWidgets.
It this case it was impossible to get it out from there since all
the mouse events were directed to the ChannelWidgets.

 * Q_OBJECT removed
 * setMeasurementVisible renamed to updateMeasurementVisibility

Signed-off-by: Christian W. Zuckschwerdt <christian@zuckschwerdt.org>

This is probably my last commit on the openhantek2 branch.
I've got it to the stage when I can see the reference signal with
 - correct offset
 - correct period
 - correct amplitude
 - correct voltage measurment label at the bottom
 - it is possible to change the sample rate and the timebase is calculated correctly

Things which are still broken:
 - software gain control is broken (and turned off in this commit)
 - with the hardware gain control, the gain dropdown shows Vols per full
   scale instead of Volts per 1 div
 - hardware gain is not initialized when application is first started
   (need to select some other gain to initialize it)
 - hardware offset is not initialized when changing the gain level
   (need to shift the slider a bit to initialize it)
 - trigger offset and position are broken
 - axes lines are still randomly broken
 - only 10kS mode currently works
 - it is impossible to change the timebase directly using the spinner

Overall it looks like the codebase is in a pretty broken state.
At this point it feels easier to start over with something working and
incrementally apply refactoring on top of it while making sure that
none of the existing logic has changed.