Set default values for segmentation widgets based on quantitative experiments

[constantinpape]

The segmentation widgets (especially for 3d segmentation, tracking and automated segmentation) have parameters for extending boxes, projection modes etc. The default values are currently based on intuition (or copying the defaults from SAM in case of the AMG). We should instead implement quantitative evaluations for the underlying segmentation and use these to set the defaults.

This would also allow us to evaluate how different models perform in these settings.

[constantinpape]

This will be enabled by #483 (can access the relevant plugins). I will then write a function that updates the values in the widget based on the model specific defaults.
For this we now also need to compile the model specific defaults. cc @anwai98

[constantinpape]

The mechanics for this are now implemented in #486. The corresponding code is here: https://github.com/computational-cell-analytics/micro-sam/blob/dev/micro_sam/sam_annotator/util.py#L622-L670