GitLab

Since we now have sub-millisecond precision in the data, we control output
width based on the fractional part so we don't lose precision

Moves from QElapsedTimer to chrono::steady_clock.
Uses frame timestamps provided by the frame if available, compensating
for drifts w.r.t. EyeRec time.

Includes combo box for tracking although we only have one tracking method
at the moment

- Updates icons, licenses, and copyright

- Post processor (in beta; there are many things that need to be fixed and
  improved)

- Integrates gaze estimate into the gaze estimation pipeline

- Already adding Eigen as it will be necessary when we merge Grip and the
  eye model

If higher priority trigger had been open, it's queue would never be empty.
Thus, if that triggers is turned off, next trigger would never kick in.

At 400x400 the contrast is really bad; playing around a bit, this seems to
be the best it gets.

Also changes default calibration method and maximum pupil size

Uses data from both eyes for a single polynomial regression

Points considered to bias the evaluation points were being used anyway.
Although for the majority of CalibMe cases this had little effect --
because the large amount of points + spread give little weight to each
point -- it could be critical if the calibration points are only
concentrated around evaluation points -- e.g., N-Point calibrations.

Also, the whole gaze estimation logic needs to be untangled.

Do median outlier identification based on the mono / binocular calibration
setting.
Pick first available method if unknown was requested.

Moving to doubles and normalizing per resolution (so now we allow for
resolution changes given that the aspect ratio remains constant).

Makes it possible to mix polynomials

If startRecording() was called instead of startRecording(double),
videoWriter would never be initialized.

Thanks David Gil for pointing that one out