Streamlining the thresholds

[danielmwatkins]

Improving the use of thresholds in IceFloeTracker.jl

The original Matlab code for IFT heavily relied on hardcoded thresholds for various steps in the process, such as the maximum distance between floe pairs for matching, area overlap, and correlation. Carlos has done a lot of work in taking these hardcoded thresholds and changing them to user-adjustable parameters. For at least a few of these thresholds, it seems to me like a simple parameterization scheme may be more effective. For example, for the maximum drift distance, there are distances assigned to 3 pre-specified time deltas. It would be simpler (and in my opinion better justified) if we were to use a physical setting like "maximum drift velocity", so the distance threshold would be a linear function of elapsed time:
$$\Delta X_{max} = U_{max} \Delta t$$
with e.g. $U_{max} = 1.5$ m/s. That's a lot faster than typical ice drift, but it's still a possible drift rate for a day if there's, say, a strong current and steady wind.

How should we approach the thresholds that are functions of floe area? What do we consider "large" and "small" floes, and can we make those categories empirical instead of subjective?