[Inference Providers] wavespeed supoort image-text-to-image type

[arabot777]

add provider image-text-to-image
test result:

[hanouticelina]

thanks @arabot777 for the PR, i'm closing this in favor of #1882.
Note that the input image should be optional for image-text-to-image and image-text-to-video. In #1882, we handle this by sending a 1x1 fully transparent image when no image is provided, rather than conditionally omitting the image field. This approach was chosen because WaveSpeed's API endpoints for image-to-image and image-to-video still expect an image parameter to be present, so we satisfy that requirement with a minimal placeholder image. What do you think about this?