fix(import): exclude rowHeader and hidden columns from xlsx/array import mapping

[simonyang08]

Fixes #4726

Root cause

Import.structureArrayToColumns maps imported cells to columns by raw index from this.table.getColumns(). That array includes:

• the rowHeader column, pushed at index 0 by ColumnManager.setColumns when rowHeader is configured,
• hidden (visible: false) columns, which stay in the list.

The Export side strips both (rowHeader has no field, and columnVisCheck drops it along with invisible columns), so an export → import round-trip shifts every value by one slot when rowHeader is configured and silently writes into hidden fields otherwise. This matches the reporter's JSFiddle and the follow-up observation that hidden columns trigger the same misalignment ("the upload is completely column position/index-based").

Fix

Filter the column list in structureArrayToColumns to skip isRowHeader and !visible columns before the index-based mapping, mirroring what the export actually wrote to the file:

columns = allColumns.filter(function(column){
    return column && !column.isRowHeader && column.visible;
});

Two regression tests added to test/unit/modules/Import.spec.js (rowHeader case + hidden-column case). Both were verified to fail on unpatched main and pass with the fix; full unit suite green (498/498).

Known boundaries (not regressions, noted for reviewer awareness)

• A visible column with no field is excluded by Export (column.field check) but still consumes an import slot; this predates this change and is outside rowHeader breaks table.import("xlsx") #4726's scope.
• The extreme config rowHeader: {field: "rowid", ...} (rowHeader with a real field) is included by Export by default but skipped by Import after this change. The issue as filed concerns the default fieldless rowHeader, so Import skips isRowHeader unconditionally; happy to adjust if maintainers prefer full symmetry there.

DCO: Signed-off-by: simonyang08 <ppt5928@gmail.com> (commit 5430fe2).