Upgrade filter logic

[jeremykubica]

The filtering done in analysis_utils.py only filters out results with < 3 observations (after sigmaG clipping) regardless of the number of observations provided in the config. This tightens the threshold to match the given number of observations. This changes the diff test by dropping out results that had too few observations.

This also combines the two filtering steps for efficiency, so we are only looping through the results once (instead of twice).

Adds two new filters:

• CombinedStatsFilter filter combines likelihood bounds and minimum number of observations. This is for efficiency
• DurationFilter filters based on the time duration of the track. This is a potential to explore in the future.

Finally the PR removes the custom _make_trajectory helper function in one of the tests and replaces it with the make_trajectory function from trajectory_utils.py.

[maxwest-uw]

sorry I'll get this done by EoD monday, I'm trying to chase down a hipscat-import bug atm 😪