test: rewrite PR sync tests to reflect purgatory-first announcement flow

The tests now correctly reflect the actual purgatory behavior:
1. Announcement goes to purgatory (StateOnly) - not immediately accepted
2. State event goes to purgatory
3. Git push promotes announcement to Full and releases state event
4. PR event is sent AFTER announcement promotion (accepted since repo is Full)
5. PR commit push releases PR event from purgatory

This matches the design: announcements require git data validation before
being promoted to the database, which means PR events can only be accepted
for repos with promoted announcements.

Also routes relay stdout to /tmp/relay-{port}.log for easier debugging.

0 files changed, 0 insertions, 0 deletions