catch Error if `model_type` is redefeined in model config

[redspottedbittern]

If either the key model_type or taskmodule_type (see here) are used as a parameter in the constructor of a model or taskmodule, respectively, it is not possible to instantiate the respective objects via .from_pretrained and this error message is displayed (dummy is the value of the redefined model_type):

Error in call to target 'pytorch_ie.auto.AutoPipeline.from_pretrained':
RegistrationError('dummy is not a registered name for PyTorchIEModel.')

[ArneBinder]

Thanks for reporting! We should either prohibit overwriting these config entries or at least show a better error message.