Tests TestSemaphoreRobust and TestSemaphoreN failed randomly

[Loyalsoldier]

See here: #68

[benja-wu]

Thx for reminding.
Let me figure out what's wrong here.

• MacOS go test failed due to the semaphore_test
• Ubuntu go test failed due to the cluster_test

[xxx7xxxx]

Try it multiple times, it will reproduce the failure. The reason here is it gave a too-small gap of the time in a limited CPU resource. So the failure will be eliminated by increasing the time gap. But my real proposal is to make it a benchmark test instead of unit test. It should not use time consumption to decide whether the unit test is successful. Github action apparently gives a very small resource of CPU, so it needs much longer time than our modern developing machines.

@qdongxu please check it out.

$ cd pkg/util/sem
$ GOMAXPROCS=1 go test -v
=== RUN   TestSemaphore0
--- PASS: TestSemaphore0 (0.10s)
=== RUN   TestSemaphoreRobust
--- PASS: TestSemaphoreRobust (1.07s)
=== RUN   TestSemaphoreN
    semaphore_test.go:153: time too long: 211.624608ms, sem: 799, trans: 800
--- FAIL: TestSemaphoreN (2.33s)
FAIL
exit status 1
FAIL    github.com/megaease/easegress/pkg/util/sem      3.624s

[qdongxu]

We implemented a Semaphore strategy when there's not an 'official` semaphore. Now we can employ https://pkg.go.dev/golang.org/x/sync/semaphore, then less unit test is required.