to_markdown() scaffold — Dataset Overview + Column Summary table

[DEVunderdog]

Parent

#268

What to build

Implement the new compact to_markdown() on StructuralProfileResult. At this slice, it produces a Markdown document with two sections: a Dataset Overview (all scalar dataset-level fields from DatasetStats — row count, column count, memory bytes, duplicate count/ratio, overall sparsity, was_chunked, row missingness distribution — but no memory_breakdown dict and no missingness_matrix) and a Column Summary table with one row per column showing name, semantic type, missing %, severity, and flags. No per-column detail sections yet.

Acceptance criteria

• to_markdown() returns a valid Markdown string with a ## Dataset Overview section containing only scalar fields (no memory_breakdown, no missingness_matrix)
• to_markdown() includes a ## Column Summary table with one row for every column in result.columns
• Each Column Summary row shows: column name, semantic type, effective null ratio (%), missingness severity, missingness flags, numeric flags
• Unit tests assert Dataset Overview contains row_count, memory_bytes, duplicate_ratio, overall_sparsity and does NOT contain per-column memory bytes or the missingness matrix
• Unit tests assert the Column Summary table has exactly one row per column in the result

Blocked by

• Blocked by to_full_markdown() — rename lossless implementation + tests #269