fix: flaky screenshot test timing - 0.2s pause raced against the command's own 0.1s timer

Confirmed via a real CI failure on Python 3.9 (v0.1.22's tag push):
test_save_screenshot_reports_failure_without_crashing failed there
despite passing locally and on other Python versions - a genuine race,
not a real product bug (the feature itself was separately confirmed
working via live testing in a real container before this release).
Both publish workflows' own separate test runs on the same commit
happened not to hit the same timing loss and passed, so v0.1.22 did
ship correctly - this is purely a test-reliability fix, nothing to
re-release for.

Increased both screenshot tests' pilot.pause() from 0.2s to 1.0s - a
10x margin over the command's own set_timer(0.1, ...) delay, verified
stable across 5 repeated local runs before pushing this time, not
just a single one.

Author: andlo

tests/test_command_palette.py

@@ -407,7 +407,14 @@ async def test_save_screenshot_uses_home_directory_not_cwd(tmp_path):
     async with app.run_test() as pilot:
         with patch.object(app, "deliver_screenshot", return_value="/home/fake/screenshot.svg") as mock_deliver:
             app._save_screenshot()
-            await pilot.pause(0.2)  # the real command schedules via set_timer(0.1, ...)
+            # Real, confirmed flakiness: 0.2s here raced against the
+            # command's own set_timer(0.1, ...) and lost on a Python
+            # 3.9 CI runner once (the timer hadn't fired yet by the
+            # time this returned) despite comfortably passing locally
+            # and on other Python versions - a 10x margin over the
+            # timer's own delay should be enough headroom that this
+            # doesn't recur even under CI load/scheduling jitter.
+            await pilot.pause(1.0)
         mock_deliver.assert_called_once_with(path=str(Path.home()))
         conv = app.query_one("#conversation", RichLog)
         text = "\n".join(str(line) for line in conv.lines)
@@ -420,7 +427,7 @@ async def test_save_screenshot_reports_failure_without_crashing(tmp_path):
     async with app.run_test() as pilot:
         with patch.object(app, "deliver_screenshot", side_effect=OSError("disk full")):
             app._save_screenshot()
-            await pilot.pause(0.2)  # must not raise
+            await pilot.pause(1.0)  # must not raise - see the sibling test's own comment on why 1.0s
         conv = app.query_one("#conversation", RichLog)
         text = "\n".join(str(line) for line in conv.lines)
         assert "failed" in text.lower()