Split responses from reasoning models into their own column in responses table

[rlarson20]

I've been toying around with llm and datasette, mostly with models with reasoning capabilities. I have a feeling that it'd be more difficult than simply checking if the response starts with or something similar, but the idea seems solid: in addition to the response column in the responses table, add a 'reasoning_response' and 'model_response' (or something with a better name than that). While this might just as easily be solved on the user side with some clever scripting, it seems like it would get use from those who are accessing the log database (which to be fair is one of the main draws of this tool IMO).