Android/iOS native core integration

[sargunv]

#544 added basic native core integration for desktop. Once that matures and catches up to our existing MapLibre Android/iOS wrappers, we should look at using our own native core integration for Android and iOS too, to eliminate API differences between the different SDKs.

• Do this only after Desktop target feature parity #570 is complete.
• The maplibre-native-bindings module should become a KMP module with Android and Native (iOS) targets in addition to the existing JVM target (for naming, see "desktopMain" -> "jvmMain" #549)
• For Android, we should be able to re-use most of the maplibre-native-bindings-jni module, except integrating with an Android Surface instead of an AWT Canvas. But rather than publishing the C++ natives as their own artifact, they should be integrated in to the .aar for maplibre-native-bindings.
• For iOS, we'll need new Kotlin/Native cinterop bindings to MapLibre Native. We can do this through ObjC (for apple platforms) or C (for all native platforms). Compose Multiplatform today only supports Apple platforms for Native, but there's other folks working on Linux and probably Windows, so it's worth trying to write a C ABI.
• If we get a working C ABI for native, it's worth investigating whether we can use that to streamline or replace the JNI integration. But we have working JNI right now, so I wouldn't abandon that to build the C one first.

Tldr: order of operations is:

• Get desktop (JVM) to parity with mobile
• Add Android support to our JNI bindings
• Add iOS support via (hopefully) C bindings
• Investigate porting JVM and Android to the same C bindings