Improve openai image model cost calculation

[reynaldichernando]

this is similar to gemini issue #2348

openai also charge based on input and output tokens

input tokens:

• the text prompt length

output tokens:

• actual image
• reasoning tokens

note:

• i didn't put image as input here, because our txt2img API doesn't allow yet for image input
• although this is supported in their API, maybe in another issue if we want to do this
• https://platform.openai.com/docs/api-reference/images/createEdit
• gpt-image-1.5 is multimodal, which allows it to output text, so same concern if it outputs text token, and it is unnaccounted

https://openai.com/api/pricing/