Align UR2CUTE integration with upstream release (auto-threshold, masked losses, tests)

[FH-Prevail]

Summary

• mirror the latest standalone UR2CUTE estimator: wire BaseModel knobs, enable padding/sanity overrides, and add auto-threshold bookkeeping + masked BCE/regression losses
• tighten inference by using the learned threshold and exposing a _return_components flag so training/validation can consume both CNN outputs safely
• skip validation when val_size == 0 to keep Lightning happy on short datasets and add pytest smoke coverage (fit/predict + auto-threshold)

Testing

• python -m pytest tests/test_models/test_ur2cute.py -q --cov-fail-under=0

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you all sign our Contributor License Agreement before we can accept your contribution.
1 out of 2 committers have signed the CLA.

✅ FH-Prevail
❌ sina

---

sina seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[FH-Prevail]

Hi together,
What are the steps I should do at this moment?
for detail about the method , I already published it in a journal here:
https://www.mendel-journal.org/index.php/mendel/article/view/413