fix(excel): flush import data to disk and surface write failures

[Jeremy6660]

Summary

officecli import <xlsx> /SheetN <file.csv> reports success (Imported N rows x M cols into /Sheet1 ...) but writes nothing to the sheet. The xlsx on disk stays empty, so the imported data is silently lost — CSV and TSV both reproduce this.

Root cause

ExcelHandler.Import writes rows/cells into ExcelHandler's in-memory package, but:

1. SaveWorksheet enqueued the part into _dirtyWorksheets without setting the Modified flag. ExcelHandler.Dispose() therefore treated the whole session as read-only and discarded the in-memory copy (reverted to the on-disk bytes) — the _dirtyWorksheets entries were never flushed.
2. The import command layer (CommandBuilder.Import) never explicitly saved; it relied on Dispose(), which did nothing because of (1), and reported the returned message string as success regardless.

The set path worked because it goes through the resident/save machinery that does set Modified.

Fix

• ExcelHandler.Helpers.Sheet.cs: SaveWorksheet now sets Modified = true — a dirty worksheet is itself proof the package was modified, regardless of which caller enqueues it. This is the root-cause fix.
• CommandBuilder.Import.cs: after handler.Import(...), call handler.Save() explicitly so write-back happens before the success message is printed. Any write-back failure now surfaces as a command error instead of being swallowed on Dispose().

Validation method (per CONTRIBUTING Rule 2)

Before my fix (v1.0.143 baseline):

officecli create t.xlsx
officecli import t.xlsx /Sheet1 data.csv   # → "Imported 3 rows x 2 cols into /Sheet1 starting at A1"
officecli view t.xlsx text                 # → (empty)  ❌ data not present
unzip -p t.xlsx xl/worksheets/sheet1.xml | grep alice   # → (empty)  ❌ on-disk XML has no data

After my fix:

officecli create t.xlsx
officecli import t.xlsx /Sheet1 data.csv   # → "Imported 3 rows x 2 cols into /Sheet1 starting at A1"
officecli view t.xlsx text
# → [/Sheet1/row[1]] A1=name B1=score
# → [/Sheet1/row[2]] A2=alice B2=88      ✓ data present
unzip -p t.xlsx xl/worksheets/sheet1.xml | grep alice   # → present   ✓ on-disk XML has data

Also cross-verified with openpyxl (independent of officecli's own readers): the workbook loaded by a third-party library contains the imported rows after the fix, confirming the bytes were really persisted to disk.

Note: I validated by building the patched source with dotnet build and running the CLI against the compiled officecli.dll (macOS arm64, net10.0.302). CI's multi-platform dotnet build matrix should confirm the same for the other RIDs.

Checklist

• One atomic change (import data persistence)
• Compiles clean (dotnet build, 0 errors)
• Verified data persists to disk (officecli view + unzip XML + openpyxl cross-check)