Image import and camera capture for multimodal canvas input.

## Goal

Allow users to bring real-world visual content into PenEcho by importing an existing image or taking a photo with the device camera.

## Proposed behavior

- Add an **Import image** action that accepts common formats such as PNG, JPEG, and WebP.
- Add a **Take photo** action on supported devices and browsers.
- Let the user preview, confirm, retake, or cancel a camera capture.
- Insert the imported or captured image into the current canvas while preserving its aspect ratio and fitting it within the visible canvas.
- Allow the image to be selected, moved, resized, and deleted using the existing canvas interactions.
- Include visible imported images in the canvas content sent to the configured AI workflow.

## Permissions and privacy

- Request camera permission only after an explicit user action.
- Never activate the camera in the background.
- Handle unsupported browsers and denied camera permissions with a clear message.
- Keep the image local until the user explicitly triggers an AI request.
- Avoid retaining unnecessary image metadata.

## Acceptance criteria

- A user can import a PNG, JPEG, or WebP image from local storage.
- A user can capture a photo on a supported phone, tablet, or computer.
- The imported or captured image appears correctly on the canvas without distortion.
- Camera denial, cancellation, unsupported devices, and invalid files are handled gracefully.
- The feature works with the existing canvas selection and AI request flow.
- Automated tests...