[DWAA1660]

What and why

Fixes #110 and #112, both in the CSV reader that GridCraft read-back uses (table_from_gridcraft_csv). They touch the same lines of csv::to_table, so they're one PR in two commits rather than two PRs that would conflict.

#110: cells were read as Markdown. to_table ran every cell through markdown::parse_inline, so a spreadsheet value *literal* became italic literal and [a](b) became a link. Exporting the table again lost the characters. A cell is now one plain Span holding its value as written, or no span for an empty cell, which is the same shape the Markdown table parser gives an empty cell. The page writer already escapes literal text (#92), so these values also survive a page save.

#112: blank rows of a one-column table were dropped. The reader skipped any line whose single cell was blank. That skipped real blank rows too, and every row under them moved up. Now:

• from_table quotes a blank cell when it's the only cell on its row (""), so the line isn't empty
• rows_of records whether a line had a quoted cell, and to_table keeps such a line
• truly empty lines (and whitespace-only unquoted ones, as before) are still skipped, so trailing newlines and the nothing_in_nothing_out cases behave as they did

How it was checked

• New a_cell_that_looks_like_markdown_is_read_as_it_is_written: four inputs full of Markdown punctuation are read with no marks and written back character for character
• New a_blank_row_of_a_one_column_table_is_kept: a one-column table with an empty and a whitespace-only row round-trips exactly, and spreadsheet-style "" rows among blank lines read as expected
• Existing CSV tests unchanged and passing; cargo test -p notecraft-formats: 73 passed
• GridCraft read-back: gridcraft_csv_read_back_preserves_quoted_cells_and_adds_columns and the UI test a_table_is_read_back_from_the_spreadsheet_beside_the_page pass
• cargo fmt --all -- --check, cargo clippy -p notecraft-formats --all-targets -- -D warnings: clean

🤖 Generated with Claude Code