Refactor how we manage firebase gradle plugins for the test-app

Nobody has claimed this yet.

Following up from https://github.com/mapbox/mapbox-navigation-android/pull/4075
Other approaches considered https://github.com/mapbox/mapbox-navigation-android/issues/4071

### Problem

Right now we have an empty google-services.json to allow customers to run the app without access to Firebase. But we don't want to require an empty

This ticket is to consider approaches and find a long term solution. Iterations attempted may need to be expanded on.

This ticket may also require a request from Firebase, to give a clean recommendation.

### Requirements

- Customers outside of Mapbox should be able to clone the repo and run the test-app
- Firebase robotic runs should be able to run the test app
- Firebase robotic runs should include crashlytics

### Approaches

Pass in a gradle parameter to disable plugins ./gradlew ... -PuseFirebasePerfAndCrashlytics=true

if (useFirebasePerfAndCrashlytics) {
apply(plugin = "com.google.firebase.firebase-perf")
apply(plugin = "com.google.firebase.crashlytics")
println("Firebase performance and crashlytics plugins applied")
}

This doesn't work because the plugin is required even if the build type is not used
``` groovy
release {
minifyEnabled true
signingConfig signingConfigs.release
manifestPlaceholders = [enableCrashReporting: "true"]
firebaseCrashlytics { << == required even when the variant is debug build
mappingFileUploadEnabled = true
nativeSymbolUploadEnabled = true
strippedNativeLibsDir = 'build/intermediates/stripped_native_libs/debug/release/lib/'
unstrippedNativeLibsDir = com.mapbox.gradle.NativeDownloadTask.UNSTRIPPED_NATIVE_LIBS_PATH
}
}
```

Start by reading pull request 4075 and issue 4071, then inspect the test-app Gradle configuration and its google-services.json usage. Compare the proposed Firebase plugin approaches against the three requirements: external customers can clone and run the app, Firebase robotic runs can run it, and those runs include Crashlytics.

Written by the indexing model from the issue text.