Reduce AppHangPlugin heartbeat overheads

[lemnik]

Goal

Reduce the CPU load (and "spamminess") of the heartbeat messages posted by the AppHangPlugin.

Design

Restructured the heartbeat such that:

• Heartbeats only adjust the "last heartbeat" timestamps
• Heartbeats are triggered by the monitor thread on a half-interval to the first "app might be hanging" time (either when we start sampling, or when we report an AppHang)

The heartbeats are also now triggered after appropriate delays (and only if there is no heartbeat message already on the queue), with the delay being calculated based on whether the monitor thread considers the app to be:

• running normally (wait for a heartbeat interval)
• stuck and requiring stack sampling (wait for a stack sample interval)
• within an app hang (wait until the next heartbeat)
• after an app hang and within the cooldown (wait for the cooldown period)

The "heartbeat interval" is half of the appHangThreshold or stackSamplingThresholdMillis (if configured):

    private val heartbeatInterval =
        if (isSamplingEnabled) samplingThresholdMillis / 2
        else appHangThresholdMillis / 20

Testing

Manual testing more complex scenarios with some dedicated message queue debugging added. Relied on existing tests continuing to pass for the overall plugin logic.

[bugsnagbot]

Android notifier sizes

Format	Size impact of Bugsnag (kB)	Size impact of Bugsnag when Minified (kB)
APK	1851.29	1608.64
arm64_v8a	659.71	413.95
armeabi_v7a	598.28	352.52
x86	729.33	483.57
x86_64	700.66	459.0

Generated by 🚫 Danger

[hamudi-ahmed]

LGTM