Skip slow windows tests

[guarin]

What has changed and why?

• Skip some tests that are slow on Windows

CI currently takes nearly 30min which is too long for a single PR. This should bring down the runtime quite a bit. Those are the slowest tests right now:

-- Docs: https://docs.pytest.org/en/stable/how-to/capture-warnings.html
============================ slowest 20 durations =============================
263.12s call     tests/_commands/test_train_task.py::test_train_object_detection
252.75s call     tests/_commands/test_train_task.py::test_train_panoptic_segmentation
193.15s call     tests/_commands/test_train_task.py::test_train_panoptic_segmentation__dinov2
187.80s call     tests/_commands/test_train_task.py::test_train_instance_segmentation
121.43s call     tests/_commands/test_train_task.py::test_train_image_classification__multilabel
104.33s call     tests/_commands/test_train_task.py::test_train_image_classification__multiclass
24.59s call     tests/_commands/test_train.py::test_pretrain__checkpoint
20.63s call     tests/_commands/test_common_helpers.py::test_file_locking_concurrent_increments[10-False]
20.59s call     tests/_commands/test_common_helpers.py::test_file_locking_concurrent_decrements[5-5-True-False]
20.56s call     tests/_commands/test_common_helpers.py::test_get_dataset_temp_mmap_path__concurrent_context_managers
17.44s call     tests/_commands/test_train.py::test_pretrain__cpu
13.89s call     tests/_commands/test_common_helpers.py::test_file_locking_concurrent_decrements[10-3-False-False]
13.46s call     tests/_commands/test_embed.py::test_embed__cpu
3.87s call     tests/_methods/distillation/test_distillation.py::TestDistillation::test_teacher_parameters_are_frozen
3.77s call     tests/_methods/distillationv2/test_distillationv2.py::TestDistillationV2::test_teacher_parameters_are_frozen
3.71s call     tests/_methods/distillationv2/test_distillationv2.py::TestDistillationV2::test_distillation_configure_optimizers_lr_scaling[128-2.598076211353316]
3.71s call     tests/_methods/distillationv2/test_distillationv2.py::TestDistillationV2::test_distillation_configure_optimizers_lr_scaling[768-6.3639610306789285]
3.65s call     tests/_methods/distillationv2/test_distillationv2.py::TestDistillationV2::test_distillation_configure_optimizers_lr_scaling[1536-9.0]
3.63s call     tests/_loggers/test_logger_helpers.py::test_get_callbacks__mlflow_user_config
3.48s call     tests/_methods/distillationv2/test_distillationv2.py::TestDistillationV2::test_distillation_configure_optimizers_lr_scaling[384-4.5]
======== 2549 passed, 1150 skipped, 520 warnings in 1424.26s (0:23:44) ========

I kept object detection because it is the most important model.

How has it been tested?

(Delete this: Please describe the tests that you ran to verify your changes. Provide
instructions so we can reproduce. Please also list any relevant details for your test
configuration.)

Did you update CHANGELOG.md?

• Yes
• Not needed (internal change)

Did you update the documentation?

• Yes
• Not needed (internal change without effects for user)

[guarin]

/review

[Copilot]

Pull request overview

This PR reduces Windows CI runtime by skipping selected long-running integration tests in the training-task command test suite.

Changes:

• Skip instance segmentation, panoptic segmentation (including dinov2 variant), and DICOM semantic segmentation training-task tests on Windows.
• Clarify the Windows skip reason string for the DICOM semantic segmentation test.