Windows event log tests regression in GitHub windows-2022 test runner image

[SaxyPandaBear]

Windows event log unit tests are failing consistently in GitHub Actions.
Example below:

--- FAIL: TestReadWithBothSources (2.04s)
    wineventlog_test.go:122: seekToEnd() current count 0
    wineventlog_test.go:128: seekToEnd() total 0
    wineventlog_test.go:175: 
        	Error Trace:	wineventlog_test.go:175
        	            				wineventlog_test.go:110
        	Error:      	Not equal: 
        	            	expected: 10
        	            	actual  : 0
        	Test:       	TestReadWithBothSources
Error:  	Messages:   	expected [Application] [ERROR] [777] [CWA_UnitTest111] , 10, actual 0
FAIL

I think that this is attributed to a recent change to the Windows OS that GitHub is using for the runner: actions/runner-images#4856

What corroborates that these failures are related to the GitHub runner change is that there are recent workflows that succeeded.

https://github.com/aws/amazon-cloudwatch-agent/runs/5009246587?check_suite_focus=true is a recent PR run that succeeded, which uses the Windows Server 2019 image.

https://github.com/aws/amazon-cloudwatch-agent/runs/5030533176?check_suite_focus=true is a recent failed run, which uses the Windows Server 2022 image. Though not 100% sure the issue, it's the only real change that would have affected the tests from what I can tell.

I think what should be done is:

1. Update the workflow right now to use the windows-2019 image for the test runner - the linked GitHub issue indicates that Windows Server 2019 is still supported, but not the default.
2. Triage the test failures in a separate, later PR on Windows Server 2022.