fix(inference/fal-ai): b64 response format argument

[badayvedat]

fal-ai endpoints are using sync_mode: true argument to receive the result as a base64 data uri
here is an example input/output with the parameter: https://fal.ai/models/fal-ai/flux/dev?share=c1793354-3994-44ef-a50b-31ef2ee9c90f

[julien-c]

launching the CI to see if it passes the existing fal tests

[SBrandeis]

Code looks good to me!

@badayvedat can you update the tapes.json file, please? It's used to replay HTTP responses in the CI.
You can do so by running test with VCR_MODE=cache:

VCR_MODE=cache pnpm run test

[badayvedat]

thanks @SBrandeis ! updated them
fyi the new response objects are base64 images - which increased the file size of tapes.json significantly -

[SBrandeis]

Nice, thank you @badayvedat !

[badayvedat]

thanks for looking into it and finalizing @SBrandeis !