OCR PPTX converter crashes on SVG images without a raster fallback

Summary

PptxConverterWithOCR still crashes on SVG pictures that do not have a raster fallback, even though the core PptxConverter was fixed for this in #2233.

Current OCR path:

if self._is_picture(shape):
    image_stream = io.BytesIO(shape.image.blob)

For an SVG-only picture, python-pptx raises ValueError("no embedded image") when shape.image is accessed. The core converter now resolves the <asvg:svgBlip> relationship directly, but the OCR converter does not use that helper/path.

Reproduction on current main

I used the repository's existing regression fixture:

packages/markitdown/tests/test_files/test_svg_no_fallback.pptx

In a clean local venv with the source packages installed:

core 57
ocr error ValueError 'no embedded image'

So the same file that proves #2233 fixed the core converter still fails through PptxConverterWithOCR.

Proposed scope

If this is available to take, I can send a focused OCR-only follow-up that reuses the core converter's SVG image-resolution behavior where practical and adds an OCR-side regression test using the existing fixture. No unrelated PPTX refactor.

I searched open and closed issues/PRs for an OCR-specific SVG/no-raster-fallback fix and found no matching implementation.

AI-assisted audit disclosure: found by comparing recent merged core-converter fixes against sibling OCR converters; reproduction was executed locally against current source.

Start at the PptxConverterWithOCR picture-handling path and compare it with the core PptxConverter SVG resolution introduced for #2233. Run packages/markitdown/tests/test_files/test_svg_no_fallback.pptx through the existing regression-test setup, then add an OCR-side regression test showing the fixture converts without the ValueError.