Specify AI pricing as price per million tokens

[simonhammes]

Most AI providers (e.g. OpenAI, Azure AI) publish their prices per million tokens.

In order to be consistent with this and avoid confusing system administrators, we strongly suggest that SeaTable also uses 1 million tokens as the unit of measurement when calculating the cost of AI usage.

This means updating the code inside ai_stats_worker.py to use {input,output}_tokens_1m instead of _1k.

In addition, the proposed documentation changes would need to be updated: seatable/seatable-admin-docs#273

What is your stance on this topic?