PPTX chart conversion drops parent category labels

PowerPoint charts with hierarchical category axes lose every parent label during conversion. A chart grouped by year and quarter becomes a table with repeated Q1 rows, making it impossible to tell which year's values each row describes.

Reproducer on main (eb31b5c), with markitdown[pptx] installed:

import io
from pptx import Presentation
from pptx.chart.data import CategoryChartData
from pptx.enum.chart import XL_CHART_TYPE
from pptx.util import Inches
from markitdown import MarkItDown

prs = Presentation()
slide = prs.slides.add_slide(prs.slide_layouts[6])
data = CategoryChartData()
for year in ("2024", "2025"):
    data.add_category(year).add_sub_category("Q1")
data.add_series("Sales", (10, 20))
slide.shapes.add_chart(
    XL_CHART_TYPE.COLUMN_CLUSTERED,
    Inches(1), Inches(1), Inches(8), Inches(5), data,
)
stream = io.BytesIO()
prs.save(stream)
stream.seek(0)
print(MarkItDown().convert_stream(stream, file_extension=".pptx").markdown)

Actual table:

| Category | Sales |
|---|---|
| Q1 | 10.0 |
| Q1 | 20.0 |

Expected: retain both 2024 / Q1 and 2025 / Q1 (or otherwise include the parent labels), with each row's values unchanged.

PptxConverter iterates the leaf categories. python-pptx exposes the complete parent-to-child paths through Categories.flattened_labels. This also reproduces with three category levels. I have a focused fix and in-memory PPTX regression tests ready.

Start in the PptxConverter code where leaf categories are iterated, then inspect python-pptx's Categories.flattened_labels and the existing in-memory PPTX tests. Reproduce the two-level and three-level cases from the issue, preserve the parent-to-child labels in the Markdown table, and verify that each row's values remain unchanged.