- Introduces `ExtrinsicParameters`-Struct
- Extract `ExtrinsicBox` from `Control`

- Move slots to FilterBeforeBox when concerning its elemets
- Fix Tab order of new extracted widgets
- Delete dependency to Control by AutoCalib

- Also add -autoIntrinsic documentation

- Move "filter before" GroupBox into own widget
- Move "intrinsic" GroupBox into own widget
- Save intrinsic calibration in struct
- Use said struct as parameter for `CalibFilter`
- Add `setValue`-helper function for loading values into UI respecting min/max
- Move PeTrack-Version from Setter to Constructor of Petrack-object
- Add Unit-Tests

- Changes shortcut for jumping to frame

This
 - enables editing the .ui-file without changing control.h; reducing recompile time
 - encapsulates the widgets themselves from everything outside of Control

The custom spinboxes ignore the wheel event, i.e. can not be changed with the mouse wheel.

This MR does not touch sliders. **Sliders can still be changed by scrolling** and spinboxes connected to sliders (i.e. translation of coordinate system) are changed as well, when the slider is changed by scrolling.

- Preparation on a switch to MSVC on windows
  - Reason: Support for vcpkg, while MinGW is only community supported

- Alt + holding left mouse-button enables to move the point under the cursor by dragging it with the mouse
- Point isn't moved anymore as soon as either the mouse button or alt are not pressed anymore
- Added documentation in "Key Bindings" Window for this and for !188 (Jumping to frame of trackpoint under cursor; Alt + dbl-Click)

Add AutoSave mechanism. The manual changes to the trajectories are counted. After 10 changes, the trajectories are exported to a hidden file. This hidden file can be loaded in the case of an crash of PeTrack. The project file is saved time based.

Enables to move to a frame by clicking on a trajectory via Alt + Dbl-LClick

- Add large comment describing the approach taken
- Use OpenCV data types when possible to do explicit matrix-vector-calculations

We previously used a comparison to quality 100 for checking, if a point is a detection or not in the changed code. This was replaced by `80`, so a color-detection on e.g. a multicolormarker with code, which has quality 90, still gets counted as detection.

Also a helper function `isDetection` was created, so that this knowledge about the quality <-> detection relationship isn't scattered everywhere in the code.

There was an error, where a person, who was already deleted because they were merged with another, still got evaluated. During this evaluation, the person which got the index of the deleted one got processed, or a segmentation fault happened, when the deleted person was the last one.

This is fixed by setting `mStatus` to not only signal not being tracked, but also, already being merged away.