feat(excel): range style presets — set '/Sheet1/B2:T6' --prop stylepreset=…

[ylz92871-dotcom]

feat(excel): range style presets — set '/Sheet1/B2:T6' --prop stylepreset=…

What: six one-command presets for free-form ranges — table_header,
table_banded, kpi_card, metric_positive, metric_negative, note_gray. Each
preset is a bundle of EXISTING cell style props (fill, font.color, bold,
size, italic, alignment.horizontal) expanded through the same
ApplyCellProperties path a hand-written set uses ("expand, don't invent" —
mirrors WordHandler.StylePreset from #376), so nothing a preset requests
can render as something else. Accent-keyed colors (ACCENT1/ACCENT1_TINT)
resolve from the workbook theme part at apply time with the Office default
fallback (4472C4) when no theme part exists. table_banded stripes the odd
range-rows and refuses ranges taller than 1000 rows, pointing at
add --type table for native banding. Unknown presets fail listing the
available names; cell-anchored stylepreset hard-rejects pointing at the
range form. Explicit cellProps in the same call apply first — the preset
wins conflicts deterministically.

Why: dashboards and reports need the same handful of range looks (header
band, zebra rows, KPI cards, good/bad metrics, muted notes) — hand-writing
five style props per range is error-prone and agents drift between
conventions. Pure sugar: the expansion adds no new style semantics, so
the break radius is zero.

How to verify:
officecli set data.xlsx '/Sheet1/B2:T6' --prop stylepreset=kpi_card
officecli set data.xlsx '/Sheet1/A2:A41' --prop stylepreset=table_banded
(unknown name → invalid_value with the full preset list)
Build 0 errors; local harness T-12 23 assertions green — including the
equivalence contract (preset expansion attribute-identical to the
hand-written primitive sequence, tint recomputed from the theme readback)
plus banded stripe/cap/error paths and the semantic color pairs.
Regression moat 26/26.