Fix flaky CI: default SyncOptions, pause test race, job timeout

[Menelion]

Summary

Fixes the two intermittent CI failures seen on the recent dependency PRs (#115, #117). Neither was caused by the dependency bumps.

1. Options-less sync ignored SyncOptions defaults (library bug)

SyncEngine stored the options argument as-is, so calling SynchronizeAsync() / SyncFolderAsync() / SyncFilesAsync() without options left _currentOptions null. Every _currentOptions?.X check then fell back to "off", which silently disabled the documented defaults PreserveTimestamps = true and PreservePermissions = true.

This is what made SynchronizeAsync_UpdateExistingFalse_SkipsModifications flaky on macOS: the options-less initial sync left copied files with fresh write timestamps, while the database recorded the scanned timestamps. When the runner took more than the 2-second change-detection window, the next sync saw the remote side as modified too, raised a BothModified conflict, and the UseLocal resolver uploaded, bypassing UpdateExisting = false.

Fix: _currentOptions = options ?? new SyncOptions(); at all three entry points. All other option checks keep their current behaviour, because their defaults match the old null fallback. A regression test (SynchronizeAsync_NoOptions_PreservesTimestampsByDefault) fails without the fix and passes with it.

2. Race in PauseAsync_CalledMultipleTimes_IsIdempotent (test bug, hung ubuntu jobs)

The test spawned a Task.Run per progress event, each calling PauseAsync(), then called ResumeAsync() once. A queued pause could land after that resume, leaving the engine paused with nobody to resume it. All three 6-hour hangs (runs 32477251497, 32477889169, 33693463535) stopped at exactly this test.

Fix: all pause calls now happen synchronously in the first non-scanning progress event, so they always precede the resume. The test waits for either the pause point or sync completion, and caps the final wait at 30 s so a regression fails instead of hanging.

3. CI safety net

timeout-minutes: 30 on the build job, so any future hang fails in 30 minutes instead of 6 hours.

Testing

• dotnet format --verify-no-changes: clean
• dotnet test: 819 passed, 130 skipped (integration), 0 failed
• The three affected tests passed 10/10 consecutive runs locally

Follow-up (not in this PR)

A file that exists on both sides with no sync state produces a concurrent Upload and Download of the same path during the first sync. Content is identical so it's harmless today, but it's a read/write race and should become a single compare-then-record step.

https://claude.ai/code/session_01XugGo6npZqDT15B3EcMkDk

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 78.82%. Comparing base (40b3582) to head (6d74f84).

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #119      +/-   ##
==========================================
- Coverage   78.99%   78.82%   -0.17%     
==========================================
  Files          44       44              
  Lines        4874     4874              
  Branches      723      726       +3     
==========================================
- Hits         3850     3842       -8     
- Misses        774      775       +1     
- Partials      250      257       +7     

☔ 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.