Clustering cleanup

[jeremykubica]

This PR overhauls how the clustering options and configurations to make them less dependent on the input data. Changes include:

• The "all" clustering type now uses unnormalized position and velocity (x0, y0, vx, vy) instead of normalized position, velocity magnitude, and velocity angle. This means clustering eps does not change with image size.
• Drop the normalization option from all searches. The clustering should not depend on image size.
• Add an explicit "cluster_v_scale" parameter to control the tradeoff of position vs velocity instead of using the search parameter range. This is more flexible and stable over different searches.
• Rename eps to the more meaningful cluster_eps.
• Set cluster_eps default to 20.0 to account for the lack of normalization in the searches.
• Remove broken benchmark for clustering.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[DinoBektesevic]

I think this is worth it just for the name changes, but I've a question regarding the velocity_scale == 0 situation.