fix: resolve all 9 remaining XLSX parsing failures

[developer0hye]

Summary

• Update umya-spreadsheet fork to fix 9 remaining XLSX parsing failures documented in upstream: umya-spreadsheet panics and parse failures on valid XLSX files #83
• All 21 previously-panicking files now produce valid PDFs (was 12/21)
• Merge remaining_panic_files_return_clean_errors test into previously_panicking_files_now_convert

Root causes fixed in umya-spreadsheet fork

A. Backslash zip path separators (3 files: tdf131575, tdf76115, 49609)

• Add zip_by_name() helper that normalizes path separators and case
• Handle non-standard namespace prefixes (e.g. d3p1:id instead of r:id)

B. Missing optional xl/styles.xml (3 files: 56278, tdf121887, 59021)

• Apply FileNotFound graceful handling pattern (same as shared_strings.rs)

C. Arithmetic overflow & unwrap on None (3 files)

• functions-excel-2010.xlsx: Use saturating_add/saturating_sub in coordinate math
• FormulaEvalTestData_Copy.xlsx: Widen i32 to i64 with .min(18) cap
• 64450.xlsx: Guard raw_relationship.rs attributes with let-else

Test plan

• All 97 tests pass in umya-spreadsheet fork
• All 117 xlsx fixture tests pass in office2pdf
• Full workspace tests pass (excluding unrelated perf timing tests)
• No regressions in existing XLSX tests

Related: #83

🤖 Generated with Claude Code