add discrete and continuous algorithms for the big 5 behaviors(omega turn, reversal, pirouette, pause, crawl)

[ghost]

This involves several sub-goals:

• generalizing posture_probability to create 4 by N heatmaps to see what the worm is doing from this perspective. Posture probabilities can become n-gram probabilities.
• first thing to do would be to add the continuous versions of these algorithms and then see how well discrete analogues can perform.

Thinking of pauses, another interesting idea would be to look at the binary time series of active vs. inactive states. This could be the first analysis.