feat(neon_framework): queue for request manager to limit parallel req…

[vauvenal5]

Limits the amount of parallel requests to the server to not overload it with preview requests when opening a big gallery and scrolling through it.

As agreed in #1124 this is a first step.

[provokateurin]

Cool, thanks!

[provokateurin]

Can you check the test results and try to reproduce locally? I think they could be related.

[vauvenal5]

I have to look deeper into this but I believe it is a timing issue. The bigger the delay in the test the more refresh calls pass and the expected amount does not match anymore. I guess my changes lead the more calls passing but not sure yet of the details.

[vauvenal5]

Okay I believe I have narrowed it down. There are actually 5 events which call talkBloc.updateRoom not 4.

1. BehaviorSubject.seeded(Result.success(room))
   2+3) From the refresh in the constructor
   4+5) From the refresh in the test

Due to the timing it does not show up in the test until you set a sleep also after the the refresh call in the test:

// The delay is necessary to avoid a race condition with loading twice at the same time
await Future<void>.delayed(const Duration(milliseconds: 1));
await roomBloc.refresh();
await Future<void>.delayed(const Duration(milliseconds: 1));

verify(() => talkBloc.updateRoom(any())).called(4);

My change seems to every slightly switch timing making all calls finish before the verify gets triggered in the test.

I refactored the test a bit making the behavior more deterministic.