Add max input/output tokens for providers on models

[richmailYT]

Although the models themselves may be advertised as supporting 128k input/output tokens, the providers themselves may not support it.

I propose that those metrics be added to give users more information on the trade offs between using the model from one provider compared to another besides just price and throughput.