Cell "padding": be honest about indent, don't invent unroundtrippable padding

Cluster: styles

Scenario

Users formatting information-dense spreadsheets want breathing room inside cells, space between the cell border and its text, like CSS padding in HTML tables, and ask for a "cell padding" setting. The target format (OOXML) has no general four-sided per-cell padding, so a CSS-style padding API would be a value that cannot round-trip: it would either be silently dropped on write or faked in a way that does not survive reopening in Excel.

Spec note, not a corpus case: this is an API-design and honesty question, not a failing behavior. The durable value is the stance, which is to expose the format's real inset mechanism and not invent one that lies about round-trip fidelity.

Format reality

• OOXML cell alignment exposes an integer indent (0 to about 250) that offsets the text of left- or right-aligned cells by whole indent levels. This is the only native inset, and it is one-sided (leading edge only), not four-sided padding. The reference implementation already exposes alignment.indent.
• There is no native top, bottom or right padding in pixels or points. Vertical spacing comes only from row height; trailing-side horizontal spacing only from column width.
• Google Sheets and HTML export layers sometimes fake padding with indent or spaces; those do not survive as true padding in .xlsx.

Desired stance

• Do not introduce a CSS-style four-sided padding on the cell model. It cannot round-trip and would mislead authors into expecting fidelity the format cannot provide.
• Expose the real mechanism clearly: alignment.indent for leading-edge inset, documented as one-sided and level-based, with row height and column width as the axes for the other directions.
• If an ergonomic helper is wanted, it should map explicitly onto indent, row height and column width and document that it is an approximation, not a faithful padding model.

Open questions

• Is an approximating helper worth the confusion, or is documenting the native knobs (indent, row height, column width) sufficient?
• Should a set-padding attempt throw or warn, an honest rejection, rather than silently degrade to indent?

Related: set-style-over-cell-range, alignment-does-not-leak-across-cells, worksheet-to-html-export.