Interrupt prediction and tiled prediction

[jdeschamps]

Problem

In the UI, we need to be able to interrupt long processes. For instance, Lightning has a mechanism for interrupting training, which we use in the CAREamist:

careamics/src/careamics/careamist.py

Line 265 in 9a027a5

def stop_training(self) -> None:

Now, that is not the case for prediction. We probably will need to implement it directly into CAREamics.

Potential solutions

Three cases:

• Predict to disk: easy enough, we can interrupt the loop.
• Predict from tiles: we need to interrupt the Dataset (pretend that there are no more tiles) and detect that not all tiles have been created (missing last_tile?)
• Predict on a single image: probably nothing we can do?

This is linked to: CAREamics/careamics-napari#18 (comment)

@melisande-c