RFC: row/column hiding (`zc` / `zo` / `zR`-style folds)

[garritfra]

Why it's harder than vim folds

Folds in a spreadsheet are different from vim folds because:

1. Addressing must stay stable. If row 5 is hidden, A6 still refers to the same data. Folds are purely a display concern, not a model concern. The core sheet must be untouched.
2. Copy/paste interaction is non-obvious. Should yy on a folded range yank visible-only or every-row? sc-im picks visible-only; we should too, but it's a real gotcha worth documenting.
3. Selection across a fold. Visual mode j over a fold should skip the hidden rows.
4. CSV/.cell write. Hidden rows must still serialize; folds are a view, not a filter.

Proposed scope

• App.hidden_rows: BTreeSet<usize>, same for columns.
• Commands: :hide row N, :hide col LABEL, :unhide all, :unhide row N.
• Vim-style normal-mode bindings: zc hides the current row, zC hides the current column, zo reveals the nearest hidden neighbor, zR reveals all.
• Renderer skips hidden rows/cols and draws a thin marker between visible neighbors.
• j/k skip hidden rows; h/l skip hidden columns. Same for visual selection extension.
• Yank/paste/save respect hidden state per the rules above (yank visible-only, save everything).

Open questions

• Persistence. Should folds survive :e? Vim's foldmethod=manual loses them; we probably should too for v1.
• Range hide. :hide rows 5-10? Easy and useful — include in v1.
• Outline-style multi-level folds. Out of scope; if anyone wants them, separate RFC.

Out of scope (v1)

• Persistent folds across :e.
• Outline-style multi-level folds.
• Auto-hide-empty-rows.