Several fixes and improvements enabling greater applicability of the algorithm for very large datasets

[hdaan]

This pull requests contains several commits containing the following changes:

• If there are already precalculated objects in the SpatiotemporalAnalysis object, the algorithm still goes on to detect more objects instead of only returning precalculated. This is done here
• Adds intermediate saving of 4D-OBCs. A parameter is added (intermediate_saving) which allows to save objects every _n_th seed. This also enables the splitting of the 4D-OBC computation between batches of seeds. These can be defined by the new parameters resume_from_seed and stop_at_seed. Thus, the computation starts and stops at the given seed numbers, starting at 1.
• Adds an extra parameter as input to enable writing the number of seeds detected after seed detection to a text file. This allows for the automated subsetting of the seeds into smaller batches. 4D-OBC computation on smaller batches of seeds can then be done in sequence, overcoming time constraints on some HPCs.
• Adds a check with respect to the height_threshold parameter. This parameter was previously given but never used. Now, only seeds are added where the absolute difference between minimum and maximum of a seed is larger than height_threshold. This is done here
• Adds a new parameter use_unfinished. If this is False, unfinished seeds, i.e., seeds that have not returned to their initial elevation are not considered as seeds for 4D-OBC segmentation, else, they are considered, as previously done.
• Adds a check in the seed_sorting_scorefunction for seeds without neighbors. Previously these would give an error through division by zero here. Now they get assigned a very high value, giving them a very low ranking. As they don't have neighbors they won't be growing into 4D-OBCs either way. This is a workaround, which should be reconsidered, in the future.

[hdaan]

Potentially solving issues #218, #314, #324

[kathapand]

Thanks for these valuable contributions to the 4D-OBC algorithm, @hdaan!
I would like to merge all the suggested changes. But I cannot review all of them and maybe they are missing in the code changes. At the moment, only the precalculated is visible, the other parameters are not found in the file. Could you please check if these are contained or something needs to be added to the PR?

[hdaan]

That is interesting. When I look to review the changes it does add them in the code.

Do you only see the result from this one commit then? Can you maybe show me what you see when you review?

[hdaan]

There is one issue I just found in the seed_sorting_scorefunction().
Here if len(neighbors < 2): has to be changed to if len(neighbors) < 2:.

Can you update this before merging? Or should I open a new pull request?