Fix MCS disconnects on aggressive OEM process managers

[areteruhiro]

Summary

Keep the long-lived MCS connection in a foreground service even when the app is exempt from battery optimizations.

Problem

On a Xiaomi device, Cloud Messaging connected briefly and then immediately returned to the disconnected state. The device log showed Xiaomi's low-memory killer terminating the persistent process within seconds:

Kill 'app.revanced.android.gms:persistent' ..., oom_score_adj 905

The current ForegroundServiceContext skips foreground-service mode when PowerManager.isIgnoringBatteryOptimizations() returns true. On this device, the battery exemption did not prevent the OEM process manager from assigning the persistent process a cached OOM score and killing it.

Once the persistent process was killed, the MCS socket disappeared and Cloud Messaging reported that it was disconnected.

Changes

• Add an opt-in forceForeground mode to ForegroundServiceContext.
• Use it when starting the MCS connection and heartbeat service.
• Preserve the existing behavior for all other callers.

Verification

• :play-services-core:assembleMapboxDefaultDebug --no-daemon --no-configuration-cache --max-workers=1
• Build completed successfully and produced the Mapbox/default debug APK.
• The change only touches ForegroundServiceContext.java and TriggerReceiver.java.