Increase model_max_length to avoid tokenizer warnings during packing

[flxst]

The config of the tokenizer employed during packing / tokenization of the data (modalities data pack_encoded_data) has an attribute "model_max_length", see here.

While this attribute is not used in any way during packing and thus doesn't have any effect on the results (i.e. the pbin file), it throws many warnings like "Token indices sequence length is longer than the specified maximum sequence length for this model (1757 > 1024)".

We should avoid those warnings by setting "model_max_length" to a very large value (e.g. in the Tokenizer object).