Fix UTF-8 XML scanning in spreadsheet parser

[naaiyy]

Summary

• normalize XML scan offsets to UTF-8 character boundaries
• advance custom number-format scanning from the matched tag
• add regression coverage for non-ASCII number formats

The fix prevents a Rust start-byte-index-not-a-char-boundary panic on real workbooks containing currency formats such as €.

Verification

• cargo test --manifest-path spreadsheet-view/Cargo.toml
• cargo clippy --workspace --all-targets --all-features -- -D warnings
• pnpm run build:viewers
• pnpm run typecheck:viewers
• pnpm run test:viewers
• pnpm run closure:viewer
• pnpm run budget:viewers
• exact production workbook opens through rebuilt Wasm