Jni debugging and actual workaround

[kaidokert]

Based on @johnxwork work here: #4353

• Lots of printf debugging

• Fixing initialization order. JNI_OnLoad needs to happen first, then StarboardBridge, and then Content

• JNI_VERSION_1_6 and JNI_VERSION_1_4 mismatch

• And finally, the actual key problem: With multiple dynamic libraries, everyone gets their own copy of the g_vm variables in JniEnvExt. The only actual way to solve this is to move those variables into a standalone library/module ( example: starboard_jnienv). This ensures all callers "see" the same instance of it. ( print out variable pointers to ensure )

• This problem is likely present elsewhere in Starboard where it uses lots of globals. Also likely not an issue in release builds where it's all one single .so