Stand-alone function for creating output CSV file

[roaldarbol]

Currently the data digest CSV file is produced as part of the prediction step. That means that if the way we produce the output changes, or the user wants a different output (e.g. detailed vs non-detailed) they currently have to do prediction over again which is not optimal. So it would be super useful to have a standalone function for producing the CSV export with the settings the user wants. That will be useful e.g. from a notebook or CLI. With #63, we could then also allow users themselves to make decisions about how they would prefer to have their centroids calculated if they disagree with our default way (or another implemented way simply works better for their data).

[horsto]

I think I see your point, but in terms of standardizing the output format for other readers (think movement etc.) and basically "any" analysis workflow that depends on a stable output format this should be as much set in stone as possible, shouldn't it?

[roaldarbol]

Not necessarily! E.g. for loading animovement formatted Parquet files, I have a PR for a movement reader that easily handles extra columns in a graceful way: neuroinformatics-unit/movement#963.

Have a look at roaldarbol#17 to see how I have currently implemented this. I have already used this branch to compute and recompute all the skimage metrics.