Making Unit Tests Faster by Using Simulated Data

[robintibor]

Right now in several unit tests we load the BCI Competition IV 2a Dataset to then test windowing serialization etc. In my view, this is an overkill and uses a lot of memory and runtime unnecessarily. We can just create smaller entirely fake data in those tests and this will speed up the unit tests quite a lot. Also makes it more appropriate for unit tests, like will not break in case anything changes about the MOABBDataset.

Affected files under test/unit_tests:

• datautil/test_serialization.py
• datasets/test_dataset.py
• `preprocessing/test_mne_preprocessor.py
• preprocessing/test_preprocess.py
• preprocessing/test_windowers.py
• samplers/test_samplers.py
• training/test_scoring.py

[bruAristimunha]

We have a util function called: https://github.com/braindecode/braindecode/blob/ab87bbba71719b122e8b764259854c80974d9d73/braindecode/util.py#L268

maybe a easy solution