feat(image): pass reference images to GPT image model for edits

Server: forward canvas/reference images to the image model so extract,
continue, and edit requests stay faithful to the source product.
- Add ImageRequestOptions.Images and route requests with inputs to the
  /v1/images/edits endpoint, via multipart upload (default) or JSON
  image_url payload, selectable through Agent.Image.InputImageTransport.
- Extract reference URLs from prompt directives, inline @image tokens,
  and chat message image/mask contents; fold them into the idempotency
  key and prepend a reference-fidelity instruction to the prompt.

Frontend:
- Shrink CanvasToast to a compact top pill.
- Paste an internal transparent clipboard node instead of re-uploading
  when it matches the clipboard image.
- Keep agent-project polling alive on stream error while streaming.

server/internal/svc/service_context.go

@@ -304,9 +304,10 @@ func newImageGenerator(c config.Config) piagent.ImageGenerator {
 	opts := c.Agent.Image
 	apiKey := resolveImageAPIKey(opts)
 	client := piagent.NewImageClient(piagent.ImageClientOptions{
-		BaseURL:        opts.BaseURL,
-		APIKey:         apiKey,
-		TimeoutSeconds: opts.TimeoutSeconds,
+		BaseURL:             opts.BaseURL,
+		APIKey:              apiKey,
+		InputImageTransport: opts.InputImageTransport,
+		TimeoutSeconds:      opts.TimeoutSeconds,
 	})
 	if client == nil {
 		return nil