[CODE IMPROVEMENT] Improve functioanlity for separator and stop tokens

[psinger]

🔧 Proposed code refactoring

Add the separator tokens as special tokens. Potentially then also add a separate setting to use the separator tokens as stop tokens. We should at least make the selection of stop tokens easier, maybe with a string list we are splitting.

This also means we need to dump the tokenizer.

So depends also on https://github.com/h2oai/h2o-llmstudio/issues/5

Motivation

Some separator tokens might currently be encoded as multiple tokens.

Also, it can be cumbersome to manually add new tokens to the list of stop tokens.