Implement token counting per message by using the returned usage stats from the LLM API backend. Anthropic sends token stats already by default. As for OpenAI, the stream needs to be first configured to return those.
Add migration for new Message attributes input_token_count, output_token_count
Add concern Message::TokenCount for Message model
Implement capturing of returned token count from Anthropic & OpenAI and set Message attributes inside the async API-specific jobs. For OpenAI, configure the stream to return usage stats
Add basic usage dropdown-menu beneath last message that shows the number of input/output tokens when clicked on the "$" sign. This is just a quick demo for the functionality and will likely be changed to something else
Implement token counting per message by using the returned usage stats from the LLM API backend. Anthropic sends token stats already by default. As for OpenAI, the stream needs to be first configured to return those.
This MR is related to #402