Memory growth cannot differ between GPU devices

[bertsky]

https://github.com/qurator-spk/eynollah/blob/ea792d1e4ac4a722770b82dc91e71f84d5beb212/qurator/eynollah/eynollah.py#L499

according to the Tensorflow documentation, it is not enough/allowed to apply the limit to only one device, it has to be all available devices.