Defined names must be able to carry a per-sheet scope

The scenario

A workbook can declare the same defined name on more than one sheet, each scoped to its own sheet. A common pattern: Sheet1 defines Input1 pointing at Sheet1!$A$1 and Sheet2 also defines Input1 pointing at Sheet2!$B$2, as two distinct, sheet-scoped names. In Excel these are independent: a formula on Sheet1 that mentions Input1 resolves to Sheet1's definition, and the same text on Sheet2 resolves to Sheet2's. Excel stores this with a localSheetId on the defined-name entry; a name with no localSheetId is workbook-global.

The limitation this must not reproduce

The legacy model keys defined names by name alone, a single map from name to ranges. It has no place to record a scope, so it cannot represent two same-named names with different sheet scopes. Assigning the name on two sheets does not error. Instead both cells are folded into one workbook-global name with two ranges (Input1 pointing at [Sheet1!$A$1, Sheet2!$B$2]). The consequences reported by users:

• The intended per-sheet scope is lost, and the surviving name is workbook-global.
• A formula meant to reference the local Input1 becomes ambiguous and can resolve to the wrong sheet's cell.
• Because there is only one entry, editing or round-tripping can drop one sheet's definition, so "only the last instance is recorded" is a real failure mode on the read/merge path even though the naive in-memory assignment happens to keep both.

Desired behaviour

• The model represents a defined name as a (name, scope) pair, where scope is either workbook-global or a specific worksheet, mirroring OOXML's localSheetId.
• Two sheets may each declare Input1 scoped to themselves without collision, and both survive read then write.
• Name resolution is scope-aware: a lookup from a worksheet prefers that sheet's scoped name, then falls back to the workbook-global name.
• The public API exposes scope on assignment and read-back, so cell.names or a defined-names view reports the scoped name that actually applies, not a merged global aggregate.

Root cause (legacy)

Defined names are stored keyed by name, as a cell-matrix keyed by the name string, with no scope dimension. There is no representation for localSheetId, so scoped names from a file are flattened into workbook-global entries on read, and same-named scoped assignments merge on write.

Open questions for the rebuild

• API shape for declaring a scoped name: worksheet.defineName(name, ref) for a sheet-scoped name against workbook.defineName(...) for a global one, or a scope option on a single call.
• How cell.names should report when both a global and a local name cover the same cell. List both, or prefer local?
• Interaction with data-validation lists that reference a named range: the scope must survive so the dropdown still resolves in the written file (see [[defined-names-tolerate-non-address-tokens]]).