Simplify service shutdown to prevent crash on screen rotation

[ecgreb]

Overview

Simplifies service shutdown process to prevent race condition that leads to a crash when the device is rotated.

Proposed Changes

Removes FusedLocationProviderServiceImpl#shutdown(). This method was previously invoked when the system called FusedLocationProviderService#onDestroy() and did two things:

1. Clear all location requests by calling LocationEngine#setRequest(null).
2. Clear all location clients from the client manager by calling LostClientManager#shutdown().

The problem is that FusedLocationProviderService#onDestroy() would be invoked asynchronously by the system after rotating the device. By the time onDestroy() fires the client application may have already re-connected the location client and/or made a new request for location updates.

If onDestroy() fires sometime between connecting the client and requesting updates it would lead to a crash similar those reported in #170.

Fortunately it seems neither of the operations in FusedLocationProviderServiceImpl#shutdown() are needed since both are already completed by the time onDestroy() is invoked by the system:

1. Anytime a location request is removed checkAllListenersPendingIntentsAndCallbacks() is called which calls LocationEngine#setRequest(null) if all listeners have been removed (a prerequisite to stopping the service).
2. In LostApiClientImpl#disconnect() the fused location service is only shutdown if there are no remaining clients so clearing the clients again when the service is destroyed is unnecessary.

Other Changes and Cleanup

• Removes public method shutdown() from ClientManager interface.
• Replaces with package private method LostClientManager#clearClients() visible for testing only.
• Adds @Override annotations to ClientManager interface methods implemented in LostClientManager.
• Shortens request intervals in MultipleLocationListenerSingleClientActivity to speed up feedback loop.

Fixes #170

[ecgreb]

@westnordost can you see if this changes fixes the crash you reported in #170 (and doesn't introduce any side effects) since your stack trace was somewhat different from what we are able to reproduce with the Lost demo app?

[ecgreb]

I built StreetComplete locally using a Lost 2.2.1-SNAPSHOT created from this branch and it does indeed seem to solve the crash introduced in version 2.2.0.