Replace Thread.sleep with CountDownLatch in tests with short sleeps (10-50ms)

[stalep]

Summary

Several test files use Thread.sleep with very short durations (10-50ms) to wait for async command execution. These are the most likely to flake on CI since the sleep may not be long enough for slower CI runners.

Proposed fix

Replace Thread.sleep + assertion with CommandExecutionListener + CountDownLatch pattern, as demonstrated in CommandExecutionListenerTest.java and the fix for AeshCommandInvocationServiceTest (commit 897d0b95).

Pattern:

CountDownLatch latch = new CountDownLatch(1);
// Register via SettingsBuilder:
.commandExecutionListener((line, result, durationMs) -> latch.countDown())

connection.read("command" + Config.getLineSeparator());
assertTrue(latch.await(5, TimeUnit.SECONDS));
// Assert output...

Files to convert (priority order by sleep duration)

File	Sleep(ms)	Occurrences
AeshCommandInputTest.java	10ms	3
AeshCommandDynamicTest.java	10ms	1
AeshScriptTest.java	20ms	1
AeshCommandOptionValidatorTest.java	20ms	1
AeshCommandNotFoundHandlerTest.java	50ms	1
AeshConverterInvocationProviderTest.java	50ms	1
AeshCommandPasteTest.java	50ms	1
ConsoleRedirectionTest.java	50ms	6
ExportCommandTest.java	50ms	1
AeshCommandRegistryTest.java	50ms	3

Notes

• Not all Thread.sleep calls wait for command execution — some wait for completion rendering or other async events. TestConnection.waitForOutputContaining() may be more appropriate for output-based waits.
• Each test should be verified individually after conversion.