Segmenting non-3-channel image with SAM2 Large yields wrong and unfriendly error

[ctrueden]

Error message in the console:

SAM2 Large experienced an error: Currently EfficientViTSAMJ only supports 1-channel (grayscale) or 3-channel (RGB, BGR, ...) 2D images.The image dimensions order should be 'xyc', first dimension width, second height and third channels.
java.lang.IllegalArgumentException: Currently EfficientViTSAMJ only supports 1-channel (grayscale) or 3-channel (RGB, BGR, ...) 2D images.The image dimensions order should be 'xyc', first dimension width, second height and third channels.
	at ai.nets.samj.models.Sam2.checkImageIsFine(Sam2.java:375)
	at ai.nets.samj.models.Sam2.setImageOfInterest(Sam2.java:278)
	at ai.nets.samj.models.AbstractSamJ.setImage(AbstractSamJ.java:297)
	at ai.nets.samj.communication.model.SAM2Large.setImage(SAM2Large.java:128)
	at ai.nets.samj.gui.ModelSelection.loadModel(ModelSelection.java:75)
	at ai.nets.samj.gui.MainGUI.lambda$loadModel$15(MainGUI.java:248)
	at java.base/java.lang.Thread.run(Unknown Source)

To reproduce:

• File > Open Samples > Neuron (5 channels)
• SAMJ Annotator > SAM2 Large model > Go!

Suggested fix is two-fold:

1. Change the language not to hardcode the name EfficientViTSAMJ which is confusing when the model is not EfficientViTSAM.
2. Rather than dumping the stack trace to the console, display an error dialog box using IJ.error("SAMJ Annotator", errorMessage);. Dump the stack trace using the SciJava LogService's debug method, so that it only pops up the console while running in debug mode.