Feature: Use current weights for prediction

[jdeschamps]

Description

Currently, we pass both model and checkpoint to trainer.predict. My understanding of the docs is that checkpoint takes precedence.

It is not clear what model is the best to use (last vs best), especially in the context of Noise2Void. In order to avoid a behaviour that has not be clarified and described, this PR removes the checkpoint parameter altogether.

In the future, we might want to revisit that. Currently, we would advocate letting users configure the checkpoint callback and then load the correct weights themselves.

• What: Remove checkpoint parameter from careamist.predict.
• Why: Clarify which weights are used (in this case, the current ones).
• How: Remove the parameter from the predict function signature, and from the call to the trainer.

Changes Made

• Modified: careamist.py.

---

Please ensure your PR meets the following requirements:

• Code builds and passes tests locally, including doctests
• New tests have been added (for bug fixes/features)
• Pre-commit passes
• PR to the documentation exists (for bug fixes / features)