Use OAuth fallback for workflow text generation

[abidlabs]

Description

Fixes workflow text-generation calls so they only use browser-side streaming when the frontend has an HF token. If no frontend token is available, the workflow executor now falls back to the backend call_model path, where Gradio can inject the logged-in user's OAuth token and use their inference provider credits.

Also adds regression coverage that an empty manual token slot does not block OAuth or local write-token fallback on the backend.

Closes: #13536

AI Disclosure

• I used AI to investigate the workflow OAuth/provider-token path, draft the code changes, and write this PR description. I self-reviewed the changed lines and ran the tests below.
• I did not use AI

Testing and Formatting Your Code

• pytest -q test/test_workflow.py -k ResolveToken
• pnpm exec vitest run --config .config/vitest.config.ts js/workflowcanvas/workflow/workflow-executor.test.ts
• bash scripts/format_backend.sh
• bash scripts/format_frontend.sh

[gradio-pr-bot]

🪼 branch checks and previews

•	Name	Status	URL
	Spaces	ready!	Spaces preview
	Storybook	ready!	Storybook preview

---

Install Gradio from this PR

pip install https://huggingface.co/buckets/gradio/pypi-previews/resolve/c933605e02a22f2976d340ff27e408715a436f8c/gradio-6.18.0-py3-none-any.whl

Install Gradio Python Client from this PR

pip install "gradio-client @ git+https://github.com/gradio-app/gradio@c933605e02a22f2976d340ff27e408715a436f8c#subdirectory=client/python"

Import Gradio JS Client from this PR via CDN

import { Client } from "https://huggingface.co/buckets/gradio/npm-previews/resolve/c933605e02a22f2976d340ff27e408715a436f8c/browser.js";