Test App listener's POST failure shifts the log of the test, so robotstatuschecker fails it #5360

Problem
A slow Test App makes an acceptance test fail even though its keywords passed.

atest/library/test_app_listener.py POSTs every Robot Framework event to the Test App with timeout=2. When the POST fails, it logs the error with logger.debug from start_test, so the message becomes the first item in the test's body, ahead of its first keyword. A test whose [Documentation] checks LOG 1.1:... then points at that message, and robotstatuschecker fails it.

Seen in
#5358, job "Test wheels on windows-latest for BrowserBatteries tests" (run 39054424650, job 117461288070): Robot reported 941 tests, 940 passed, 1 skipped, 0 failed, then DONE rc=(1). The status checker failed Test.01 Browser Management.Load Report.New Page Reports What The Stalled Page Was Still Loading with Locator '1' matches message and it cannot have child '1'. Body item 1 was:

DEBUG  Listener: Failed to POST event: start_test, id=s1-s2-s16-t2, name=New Page Reports What The Stalled Page Was Still Loading, status=None, server=localhost:52519 - ReadTimeout(... read timeout=2)

Any test with a LOG 1... check can fail this way. The cause is not in tools/ci_failures/known_causes.json yet.

To look at

• How the listener should report a failed POST without adding a message to the test body (for example not through robot.api.logger from start_test), within the decisions in atest/CONTEXT.md and ADRs 0001 to 0003.
• Whether a 2 s timeout is right for the Windows runners.
• After the fix is merged on main, record the known cause in tools/ci_failures/known_causes.json.