Preserve grid/table cell metadata when covering and let absolute jumps escape outer hiding

[OrangeX4]

Motivation

• Preserve rowspan/colspan and other cell metadata when covering paused table.cell/grid.cell so layout is not lost.
• Ensure content that can become visible earlier than its container (fn-wrappers or absolute jumps/#meanwhile) can escape outer hiding and manage its own visibility inside pause zones.

Description

• Treat absolute jumps inside cell metadata and child metadata as escapes by setting has-fn-wrapper = true when an absolute jump is encountered.
• Introduce cover-item to apply cover-fn only to a cell's body and reconstruct the original table.cell/grid.cell node so cell metadata is preserved.
• Change covered computation to map cover-item over items when all items are cells, and otherwise call cover-fn(items.sum()).
• Add a regression test file tests/issues/i369-i373-grid-regressions/test.typ covering the grid/cell covering and #meanwhile escape cases.

Testing

• Added and ran the regression test tests/issues/i369-i373-grid-regressions/test.typ, and it passed.

[zral0kh]

the test ref images for tytanic are not yet added

[zral0kh]

and can you maybe ask codex for whether it has a good idea for a syntax/solution for #327 ?

still can't think of anything there

edit:
perhaps we could introduce a simple pause-cell function and we just throw an error when it is used outside of a cell.
and during parsing we of a table/grid we store whether such a pause or a normal pause was encountered and then either cover the cells or just the content

[OrangeX4]

codex said:

For #327, I would lean toward not adding a global option that changes the meaning of normal #pause inside tables/grids. If we support this, a more explicit API like #pause-cells / #cell-pause might be clearer: normal #pause would keep the layout and cover later cells, while #pause-cells would remove later cells from the table/grid layout until they become visible. It should probably error outside direct table/grid children.

That said, I’m not fully convinced this is worth adding yet. It introduces another pause variant and more rules users need to understand, especially around auto-placement, rowspan/colspan, and nested content. The extra cognitive burden may outweigh the benefit unless we see more real-world demand for this behavior.