Triage the 17 genuine Windows test failures #272 surfaced

[Pixnop]

PR #272 put the test suite on windows-latest and the first honest run failed 28 tests. Nine were environmental (POSIX mode bits on NTFS, a CRLF-sensitive source search) and were fixed on that branch. Seventeen remain, deliberately left red rather than papered over, and each needs a verdict: test assumption, or real Windows bug in production code.

By family:

• 12 in tests/ipc/gameHandlers.test.ts around EXECUTE_GAME. Most look like fixtures using the bare "Vintagestory" file name where Windows needs "Vintagestory.exe". Two go further and diverge on session-adoption behaviour with mismatched uids; those two must be understood before anything is renamed, since adoption keying is the most sensitive logic in the launcher and a platform-conditional difference there would be a real bug.
• 2 in tests/ipc/extraction.test.ts: one assumes a case-sensitive filesystem, and one 7-Zip progress test times out at 5 s. The 7-Zip one may become moot if Drop node-7z and 7zip-bin by moving backups off zip #222 (dropping node-7z) lands first; sequence accordingly.
• 3 in tests/ipc/pathsHandlers.test.ts under CHANGE_PERMS: two missing-expected-rejection validation-order failures and one worker timeout.

The bar for closing: the Windows test job on dev is green, with every skip carrying a reason in the skipIf and every production divergence either fixed or documented as intended platform behaviour.