feat: add single-image preview node, fix text-only detection for multi-input models

[iammojogo-sudo]

Files changed

• src/areas/workflows/nodes/ImagePreviewNode.tsx — new: single-image preview node using data URLs
• src/areas/workflows/workflowRunStore.ts — fixed text-only detection for multi-input models, simplified multi-input routing, added extraParams.text
• src/areas/workflows/WorkflowsPage.tsx — registered imagePreviewNode in node palette and type helpers
• src/areas/workflows/preflight.ts — registered imagePreviewNode in preflight checks

New: ImagePreviewNode

A general single-image preview node for workflows. Reads the output file
from disk via window.electron.fs.readFileBase64() and renders as a
data: URL — avoids CSP issues and API origin dependency. Has a
pass-through image output so it can sit inline between nodes.

Fix: text-only model detection

A model is text-only only if all its declared inputs are "text".
Previously the check only looked at the first input, which incorrectly
treated multi-input models like Flux/Qwen ("inputs": ["text", "image"])
as text-only and sent a 1×1 placeholder instead of the real image.

Additional

• Simplified multi-input routing to use src.outputType instead of slot-based matching
• Added extraParams.text alongside extraParams.prompt for generator compatibility

NOTE: Includes changes also from #193 and #203

---