fix(ocr): emit valid Markdown newlines for PPTX conversion

[MontesanoDev]

Closes #2010

Summary

PptxConverterWithOCR emitted literal \n sequences instead of actual line breaks in Markdown output.

The optional LLM caption path also imported llm_caption from a module that does not exist in the markitdown_ocr package.

Changes

• Replace literal \n sequences with actual newlines in PPTX Markdown output.
• Import llm_caption from markitdown.converters._llm_caption.
• Update PPTX OCR snapshots to expect valid Markdown line breaks.
• Add regression tests for newline output and the LLM caption import path.

Testing

pytest packages/markitdown-ocr/tests/test_pptx_converter.py -v

Result: 8 passed.

[MontesanoDev]

@microsoft-github-policy-service agree