🐛 fix(markdown): keep list markers in cells

[gaborbernat]

MarkItDown's migration loses list boundaries in pipe-table cells: * [Nasdaq](/nasdaq) * DJIA becomes [Nasdaq](/nasdaq) DJIA. Using cell_blocks="html" keeps the list markup but leaves links inside raw HTML instead of Markdown links.

The C serializer keeps compact * and numbered item markers when cell_blocks="text" flattens lists in table cells. It escapes markers inside emphasis and link text so GFM keeps those frames. Ordered start and item value attributes, nested lists, and orphan items retain their boundaries.

In a matched 20-cycle release+LTO run on the 394,444-byte Wikipedia fixture, the serializer was 1.36% slower than merged main while producing 1,416 additional bytes of list markers (192,147→193,563 bytes). Fifteen of 993 corpus outputs changed. MarkItDown's exact-head full-file benchmark measured 97.41→10.67 ms against markdownify.

[codspeed]

Merging this PR will regress 1 benchmark

⚠️ Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

⚡ 1 improved benchmark
❌ 1 regressed benchmark
✅ 578 untouched benchmarks
⏩ 32 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
❌	test_feature[select-relative-sibling]	42.7 µs	45.4 µs	-5.95%
⚡	test_feature[conformance-sections-heading]	2.1 ms	1.1 ms	+87.24%

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing gaborbernat:fix/markdown-table-lists (5f51d47) with main (c671a6d)}

Footnotes

1. 32 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩