Feature: PDF attachment extraction API (embedded files + annotation positions)

## Context
Docling is adding PDF embedded file / attachment processing.
docling-parse needs an API that exposes embedded file streams and the positions of FileAttachment annotations.

## Proposed API
On the document-level parser:
- `list_embedded_files() -> list[EmbeddedFile]` where each item carries:
- `name: str`
- `mime_type: Optional[str]`
- `size: int`
- `bytes: bytes`
- `list_attachment_annotations() -> list[AttachmentAnnotation]` where each item carries:
- `page_index: int`
- `bbox: BoundingBox` (or equivalent rect struct used elsewhere in the API)
- `file_name: str` — reference to the embedded file name

## Backward compatibility
New methods; existing parser APIs unaffected.

Implementation will be opened as a PR from fork: yonikremer/docling-parse feat/attachments.

Start at the document-level parser and inspect the existing rectangle type used elsewhere in the API. Define the embedded-file and attachment-annotation result shapes, then add the two proposed methods while keeping existing parser APIs unchanged. Done means both methods expose the listed fields and attachment positions reference embedded file names.

Written by the indexing model from the issue text.