Stream WriteTo directly to writer and fix ZIP64 local file header sizes (B4)

[AdamDrewsTR]

Summary

Two related changes: (1) stream WriteTo directly instead of buffering, and (2) fix ZIP64 Local File Header size fields for entries >4GB.

Depends on: #2321 (buffered-writer enhancements — for CopyTo)

Bug Fix: ZIP64 LFH compressed/uncompressed sizes

Problem

Go's archive/zip writes the Central Directory entry with ReaderVersion=45 for entries >4GB, but the Local File Header retains version=20 and the original (truncated) 32-bit size values. Excel strictly validates that:

1. LFH version matches CD version (45 for ZIP64) ✅ already fixed by writeZip64LFH
2. LFH compressed and uncompressed sizes are 0xFFFFFFFF when ZIP64 extended info is present ❌ not fixed until now

Without setting sizes to 0xFFFFFFFF, Excel sees a mismatch between the LFH size fields and the ZIP64 extra field, and reports the workbook as corrupt.

Fix

After patching the version to 45, also set the compressed size (offset +18) and uncompressed size (offset +22) to 0xFFFFFFFF in the LFH for all ZIP64 entries. This tells readers to use the actual sizes from the ZIP64 extended information extra field that follows the filename in the LFH.

Performance: Streaming WriteTo

Problem

The current WriteTo calls WriteToBuffer(), which materializes the entire compressed ZIP archive in a bytes.Buffer. For large workbooks (e.g., a 4GB+ sheet compresses to 50-200 MB), this means 50-200 MB of peak memory just for the output buffer, on top of the input data.

Solution

WriteTo now streams the ZIP directly to the destination io.Writer using a countWriter wrapper (to track int64 bytes written). No intermediate bytes.Buffer is needed.

For encrypted files (which require post-processing the entire ZIP), a temporary file is used instead of an in-memory buffer, accessed via the new writeToWithEncryption helper.

Additional changes

• writeToZip uses CopyTo instead of Reader + io.Copy for streaming worksheets, leveraging the larger read buffers from Enhance bufferedWriter with bufio.Writer, scratch space, and configurable thresholds (B1) #2321
• writeZip64LFH is only called when zip64Entries is non-empty (avoids scanning the entire output for files with no >4GB entries)
• Added writeZip64LFHFile for file-based ZIP64 fixup (used by the encryption path)

Compatibility

All existing tests pass including TestZip64.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.61%. Comparing base (4bebb61) to head (58e252f).
⚠️ Report is 13 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #2324      +/-   ##
==========================================
+ Coverage   99.60%   99.61%   +0.01%     
==========================================
  Files          32       32              
  Lines       26791    26977     +186     
==========================================
+ Hits        26685    26873     +188     
+ Misses         55       54       -1     
+ Partials       51       50       -1     

Flag	Coverage Δ
unittests	99.61% <100.00%> (+0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[AdamDrewsTR]

This will need to be rebased after #2321 is merged.