fix: SentryAppStateManager should always stop when closing the SDK

[kevinrenskers]

📜 Description

Force stop the SentryAppStateManager when calling SentrySDK.close(), even if you started the SDK multiple times.

💡 Motivation and Context

Closes #2455

💚 How did you test it?

Multiple unit tests. We have a test for SentryAppStateManager that tests the actual unsubscribing of the notification center, and an integration test that starts the SDK twice, stops it, and checks the start count of the SentryAppStateManager, which should now be zero. This test doesn't explicitly test the notification center unsubscribing as I can't really pass the mock center here. But we know that if the start count hits zero, that we do unsubscribe, that's what the first test checks.

📝 Checklist

• I reviewed the submitted code
• I added tests to verify the changes
• I updated the docs if needed
• Review from the native team if needed
• No breaking changes

🔮 Next steps

[github-actions]

Performance metrics 🚀

	Plain	With Sentry	Diff
Startup time	1226.80 ms	1253.16 ms	26.36 ms
Size	20.75 KiB	383.37 KiB	362.62 KiB

Baseline results on branch: 8.0.0

Startup times

Revision	Plain	With Sentry	Diff
d10145a	1232.65 ms	1257.55 ms	24.90 ms
909e73a	1217.78 ms	1229.70 ms	11.92 ms
1ce879f	1258.12 ms	1260.90 ms	2.78 ms
3fdb749	1227.42 ms	1248.48 ms	21.06 ms
9f8d429	1239.57 ms	1255.12 ms	15.55 ms
9cb6c52	1201.08 ms	1211.37 ms	10.29 ms
e2cec76	1189.48 ms	1229.84 ms	40.36 ms
fb9f27a	1218.55 ms	1249.17 ms	30.62 ms
c9129b6	1231.86 ms	1270.11 ms	38.25 ms
dcac8ad	1238.82 ms	1247.80 ms	8.98 ms

App size

Revision	Plain	With Sentry	Diff
d10145a	20.75 KiB	379.12 KiB	358.36 KiB
909e73a	20.75 KiB	383.40 KiB	362.65 KiB
1ce879f	20.75 KiB	381.81 KiB	361.06 KiB
3fdb749	20.75 KiB	383.81 KiB	363.06 KiB
9f8d429	20.75 KiB	383.40 KiB	362.65 KiB
9cb6c52	20.75 KiB	382.12 KiB	361.36 KiB
e2cec76	20.75 KiB	381.81 KiB	361.06 KiB
fb9f27a	20.75 KiB	382.11 KiB	361.36 KiB
c9129b6	20.75 KiB	381.81 KiB	361.06 KiB
dcac8ad	20.75 KiB	379.11 KiB	358.36 KiB

Previous results on branch: fix/2455-appstatemanager-stop

Startup times

Revision	Plain	With Sentry	Diff
49fd801	1245.80 ms	1278.44 ms	32.64 ms
d253b96	1235.76 ms	1260.71 ms	24.95 ms
b4b7e62	1217.86 ms	1236.50 ms	18.64 ms
677447a	1226.74 ms	1241.28 ms	14.54 ms

App size

Revision	Plain	With Sentry	Diff
49fd801	20.75 KiB	383.54 KiB	362.78 KiB
d253b96	20.75 KiB	383.43 KiB	362.67 KiB
b4b7e62	20.75 KiB	383.54 KiB	362.79 KiB
677447a	20.75 KiB	383.54 KiB	362.79 KiB

[armcknight]

Just a question and suggestion but overall LGTM 👍🏻

[armcknight]

Could you include a doc comment or rename this so that stop:NO/stop:YES is more obvious what it's doing at callsites?

[kevinrenskers]

Done

[armcknight]

So if forceStop is not YES, then we require developers to correctly balance calls to start/stop? Why can't we just always reset startCount to 0?

[kevinrenskers]

There are 2 integrations that start this manager. And only if both are stopped, should this manager be stopped as well. The only exception is if the entire SDK is closed, (which doesn't call the stop method on every integration). So then we force this to be stopped.

But yeah normally they have to be balanced.