Fixed bug in biolm.py:

• The call to the test() function in the main() function did not have a model parameter, although it is required. So I changed the test function and made it so, that it is None by default, because then a model is created in the function.
• Added the "predict" and "interpret" scenarios in the _validate_split_exclusivity() function so that the splitpos spliration must be defined error does not occur.
• Fixed a bug with the regex when reading a line from the sequence TXT file. Now only the FIRST occurence of the regex gets stripped an not multiple, which can lead to the sequence column to be deleted.
• I changed the tokenizer_config_file path when doing the predict function, because right now when using the biolm with the saluki repository, the variable pretraining_required is always set to False. So the config file of the tokenizer is never loaded during the predict option. Please take a look at this, because I do not know if my changes break anything else.