v0.4.0

Added

• Camera (ROADMAP task 1): a node that zooms, pans and turns the view of
  everything added to it. Signals x/y (the point it looks at), zoom,
  rotation; center_on(node, s), fit(node, padding, s) (measures the
  node when called, so a growing layout fits at its current size) and
  reset(s). Follow a moving node with cam.x().bind(..). Nodes added to
  the view stay on screen (titles, captions, a HUD).
• Zoomed content is drawn at the final scale on both renderers: text and
  outlines stay sharp at ×10, blurs and shadows scale with the zoom
  (golden frames camera_zoom, camera_rotate_blur; GPU within ΔE 0.04).
• oxframe check does not report out_of_frame for nodes seen through a
  camera (framing a detail is the point); inspect marks them in_camera.
• Example camera: a tour of a system diagram.

Changed

• Signal::to takes anything that converts into the value: theme tokens
  (rect.fill().to(Accent, 0.3)), integers for f64, &str for text.

Breaking

• An explicit .into() in .to(..) is now ambiguous; drop it:

  // before
  title.text().to("Hello".into(), 1.0).await;
  // after
  title.text().to("Hello", 1.0).await;