Improve CI stability

[defagos]

As a developer I need a reliable CI to have a streamlined workflow. Currently some specific tests might fail.

Acceptance criteria

• CI pass reliably on the CI.

Tasks

• [x] Use CI test logs to identify tests that regularly fail.
• [x] Adjust them to improve stability (e.g. longer timeouts).

[defagos]

Flaky tests:

• PlayerItemDurationPublisherTests.testDuration
• PlayerItemTests.testURNPlaybackMP3: Consistently fails if repeated, after the 32th iteration. Possibly an issue with a resource pool being exhausted? Can we fix the issue by avoiding fetching the content from the IL?
• PlayerSkipForwardChecksTests.testCanSkipForOnDemand
• ItemBackwardNavigationTests.testCanReturnToPreviousItemOnFailedItem
• QueuePlayerSeekTests.testMultipleSeekMonotonicity
• PlayerSkipBackwardChecksTests.testCannotSkipForLive
• PlayerSkipToDefaultTests.testSkipForDvrInLiveConditions
• PlayerSkipToDefaultTests.testSkipForDvrInPastConditions
• PlayerSkipToDefaultTests.testSkipForLive
• PlayerSkipToDefaultTests.testSkipForOnDemand
• PlayerSkipForwardChecksTests.testCannotSkipForDvr
• ...

Could we maybe also mitigate this issue by reducing the number of parallel simulators on CI agents?

[waliid]

By updating the URLSession configuration from .default to .ephemeral, we considerably improve the stability of our tests but the result is not perfect! Moreover, I think it would be wiser to avoid network calls!

[defagos]

I agree, there is indeed a task to use only local test data, see #110.

[defagos]

Similar issues are reported for slow CI agents. We should try to increase the default timeout and see whether tests pass more reliably on our agents.

[defagos]

Builds are still quite unreliable, I reopen this issue.

[defagos]

The measures applied in the associated PR were not sufficient. The PR #307 was difficult to check so remaining tolerances were adjusted to finally have reliable behavior. Tests should now hopefully be stable.