Generate character images from prompts

This is the second step of the character video system. The character workflow calls its webhook once your cast is written, or you can start it from an Airtable button on a generation config record.

It sets the Generate Images Status to in progress, looks up every character linked to the config and loops through them. For each one it sends the character's image prompt and the shared background image to GPT Image 1.5 on fal.ai, checks the job every 10 seconds until it finishes, then saves the image link on the character and logs it in a character images table. When all are done it marks the status complete and calls the voices workflow.