Determine what should be the default thresholding option/value

[thao-do]

Please describe the problem that needs to be solved by this feature
Susanne & Matheus mentioned that with the release of Megaseg, we should check with Suraj to see what is the best thresholding option/value for MegaSeg, as naive users would rely on our default, i.e. recommended, setup to generate segmentations before they get to know how to adjust it.

In general, since the plugin is targeting beginner users in ML, it's very likely that our target users would not know how or why or would diligently check out the user guide when they are using our plugin, we'd like to set them up with the best course of actions through optimized default selection/value.

How important is this feature to you? (Software team will re-prioritize all tickets based on current workload)
Medium?

Describe the solution you'd like
After prediction has run and the plugin auto-switches to the Thresholding tab, the default selection in Threshold section should be what we'd recommend users to choose for the model we offered.
Currently, it already preselects "Specific value" with the value of "128" and Suraj confirmed it would work well for MegaSeg output.

Describe alternatives you've considered

Additional context

1. What we currently have might be already the solution, but in the future when we provide other models, we should double check if the current GUI allows user to select the best thresholding method for the given models and if the default preselection is still applicable for the most cases. If not, new thresholding method (e.g. new autothresholding method, new default specific value, etc.), and guidance (e.g. labeled "recommended") might be added
2. This could also mean a new feature and a new value for model: each model we provide could have a default recommended thresholding method/value saved with the model; when the user gets to the Thresholding step, the setup under Threshold section would automatically reflect this recommendation, but user can still change it if they want to. And for models the user train themselves, the default default will be used (i.e. specific value of 128) or perhaps another feature to allow them to save a recommended thresholding method to their model