Simplify/refactor things

[micha7a]

This is an issue for all things that we could do better, but we don't because it's research code ;)

• simplify generating trajectory times from distances
• get trajectory times from timestamps (instead of tango distances)
• use DF more in reconstructing trajectory
• get right trajectory type and model size without using specific trajectory
• make pandas float32 or numpy float64
• use DF for distance plots
• clean up the mess I made in the reconstruction notebook
• remove the .csv.npy files

[micha7a]

Actually it's not worth it to simplify the times from distances, because a priori we don't know how much time it will take to travel a given distance. So we would have loop anyway, just in different place. I should add comment on that in the code :/