fix wait logic in TestWaitForPodSucceeded

[balopat]

No description provided.

[dgageot]

By reading the changes, I don't really understand how it works better but I trust you.

[codecov]

Codecov Report

Merging #3414 into master will not change coverage.
The diff coverage is n/a.

[balopat]

The logic is this for each test:

main go routine: modify pod state -> sleep(update_time) -> modify pod state -> sleep(update_time) -> ...

waitfor pod go routine: wait for (pod to be in <state>) or timeout after (timeout_time)

if (update_time == timeout_time) (like before), then this will be flaky with a high probability of a race. It's like a "super impatient client that waits for the pod to get into a phase". Given in the first two test cases the desired update happens after the first sleep. Not sure how this ever worked consistently.

if update_time << timeout_time, this fixes it, it makes the "client more patient".

The third test case is a bit superfluous, we shouldn't need to "update" the state 5 times to the same thing, one initial update should be sufficient and the "client" will timeout anyway.