Re-enable TestGroup_Go and fix flaky behavior

[mauri870]

Proposed commit message

This PR re-enables the TestGroup_Go test, which was skipped in #41223. It also includes fixes to mitigate flakiness in the test logger caused by timming issues and ensures proper waiting for success conditions within the tests.

How to test this PR locally

$ cd filebeat/input/filestream/internal/task 
$ go test -c
$ go run golang.org/x/tools/cmd/stress@latest ./task.test -test.run TestGroup_Go -test.v

Related issues

• Fixes TestGroup_Go - not all goroutines finished/"task.Group: semaphore acquire failed, was the task group closed? err: context canceled #41218

[elasticmachine]

Pinging @elastic/elastic-agent-data-plane (Team:Elastic-Agent-Data-Plane)

[belimawr]

@mauri870 did you run this test with the race detector enabled?

Looking at the improvements you did, it looks like the older version of the test should have been flagged by it 🤔

[mauri870]

Looking at the improvements you did, it looks like the older version of the test should have been flagged by it 🤔

You are right, perhaps it was not a race condition per-se. I noticed that sometimes the contents of the log would not include log messages from both goroutines, even when both had finished executing already. I think the changes to the logger are still welcome tho, it is now safe to use in multiple goroutines without syncronization.