Explore postprocessing of trajectories

[sfmig]

Try to filter out non-bird trajectories / datapoints, based on

• too short
• too fast, or too big displacement between frames
• other?

Explore using a different tracker (check boxMOT, I had a go in ethology here)

Can we connect tracklets that belong to the same ID? (This could be dealt with by using a different tracker)