Feature/Persistent Context Token Counts

[neCo2]

Removed all the useless stuff from #62, now it just gives you a little token count display next to the title of the text field/at the bottom right of the modals. There's also a table in the context modal showing all token counts.

[lmg-anon]

lgtm! Thank you for the contribution!

Removed all the useless stuff from #62

That wasn't actually 100% useless; we do need a more accurate implementation of something like "n_keep" in Mikupad. Using "n_keep" is one way to do it, but I guess it would be better if we could do something like:

• Tokenize the memory tokens (including WI).
• Tokenize the prompt.
• Truncate the prompt tokens to the context length minus the memory tokens count.
• Join the memory tokens and prompt tokens as the new prompt.
• Detokenize the new prompt (or just send the already tokenized prompt if possible).

However, I guess some endpoints don't support detokenization (like Koboldcpp), or prompts as tokens (like Koboldcpp), so this may not be feasible, and in this case we would be forced to use "n_keep".