Batching tracking and combining outputs

[sjavis]

When running the trackers over longer periods of time it is very easy to run out of storage and/or memory by having to (pre)process all of the data in one go. It would be useful to add functionality to perform batching and combining of output tracks rather than requiring users to do so. This could also enable parallelism to speed up processing if desired.

Perhaps the main difficulty is what to do about tracks at the time boundaries which could be cut off or entirely ignored. The simplest approach is to do nothing. Otherwise, there could be overlap periods and then combine tracks / delete duplicates.

[jatkinson1000]

Discussed today and longer-term we should look at processing detection in batches, concatenating the outputs, and then running the stitching on this.

A rough idea would be to run just detect on smaller input files, then provide backend code to concatenate the candidate storm files before running stitch.
Some thought would need to be given to how to iterate over inputs automatically (or not?).

Suggestion for now is to add a flag to any tracks that end on the last time of the input data file.
This will be useful more generally even with batching.