Proposal: upstream native Windows support for codedb

[nsxdavid]

Hi! I have a working native Windows port of codedb and would like maintainer guidance on whether and how to upstream it.

This is a functioning Windows version: codedb builds and runs natively on Windows, MCP works, the CLI works, indexing/search paths work, and the Windows binary passes the MCP e2e suite.

Local validation on Windows:

zig build test
Build Summary: 23/23 steps succeeded
832/839 tests passed
7 skipped

python scripts/e2e_mcp_test.py --binary zig-out/bin/codedb.exe --project .
Results: 20/20 passed

The port handles the Windows-specific issues that previously blocked native support:

• Windows stack pressure from large inline buffers and command-frame merging
• POSIX-only file/path assumptions, including / separators, /tmp, and Unix-style root checks
• Windows root/path safety, including drive roots, system directories, backslash traversal, and temp paths
• HOME vs USERPROFILE handling for data directories, projects, and snapshots
• Windows process/argv behavior for CLI subprocesses
• MCP startup and tool handling on Windows
• daemon transport through Windows named pipes

The Windows-specific behavior is guarded by platform checks, and the existing non-Windows code paths are preserved.

I saw the earlier Windows issue/PR stack (#170-#184, #234). I understand that previous work became stale after the Zig 0.16 migration, and that the later tracking issues were closed as "no Windows support planned." I'm opening this as a fresh proposal because there is now a working native Windows version based on the current codebase with passing Windows validation.

The two questions I'd like maintainer guidance on are:

1. Would you consider supporting Windows now that the blocking issues have been resolved in a working implementation?
2. If yes, how would you like the PR submitted? In particular, what branch should it be based on and what target branch should it open against?