Fix service manager crash

[carriehall]

Potentially fixes crash: https://console.firebase.google.com/u/0/project/td-project-01/crashlytics/app/android:com.todoist/issues/5c94bfaef8b88c29631d145f?time=last-ninety-days&sessionEventKey=661EFB9F0356000131FD3657D1CE98D6_1937057838887752195

Not reproducible, but I suspect the flow is something like this:

• App not running in background
• User receives notification about a new comment
• User taps notification to deep link into item details > comment list
• App crashes

According to this issue, the activity might not be "fully awake" so they'll be paused immediately and then resumed.

Applications can get the process state in Activity.onResume() by calling ActivityManager.getRunningAppProcesses() and avoid starting Service if the importance level is lower than ActivityManager.RunningAppProcessInfo.IMPORTANCE_FOREGROUND. If the device hasn’t fully awake, activities would be paused immediately and eventually be resumed again after its fully awake.

I've added logs to debug this and I can see that when you go through this flow, the service is started, paused, and started again, although in my testing the app does reach the correct foreground state (importance = 100) so this PR is still a guess about a fix.

[rastislavvasko]

Sorry for the delay! I'm unsure if we should just silently ignore this, because it could mean that the audio won't play but we won't know about it. I.e. should we at least log an event?

[rastislavvasko]

Should we find our process or are we certain it's always the first one?

[carriehall]

It should be the first one as that's the foreground one. I've added that in the failure callback though so we can log it, just in case there's something else amiss.

[carriehall]

Sorry for the delay! I'm unsure if we should just silently ignore this, because it could mean that the audio won't play but we won't know about it. I.e. should we at least log an event?

Unfortunately we don't have Datadog or Logger in AudioPlayerRecorder. I've update the callbacks so that we can log from the AttachmentDelegate. If we log the fails and successful starts we can verify if the issue was the same (i.e. we get a start/stop/start) and I'll remove the logs again if so.

[rastislavvasko]

Thanks! Let's try it out!