[core][Android] Fix crashes during the deallocation of shared objects

[lukmccall]

Why

Fixes crashes during the deallocation of shared objects.

2024-04-25 17:10:51.732 10991-11072 libc++abi               dev.expo.payments                    E  terminating with uncaught exception of type std::runtime_error: Unable to retrieve jni environment. Is the thread attached?
2024-04-25 17:10:51.732 10991-11072 libc                    dev.expo.payments                    A  Fatal signal 6 (SIGABRT), code -1 (SI_QUEUE) in tid 11072 (hades), pid 10991 (v.expo.payments)

How

We attached the Hades thread to the JVM to call a method. However, this approach was insufficient as the default destructor for jni::global_ref was called outside of the added thread scope. To address this issue, I modified the way the global ref is passed to the shared object native state.

Test Plan

• NCL ✅
• double-check if video player is deallocated correctly ✅

[expo-bot]

The Pull Request introduced fingerprint changes against the base commit: 0a8c904

Fingerprint diff

[
  {
    "type": "dir",
    "filePath": "../../packages/expo-modules-core/android",
    "reasons": [
      "expoAutolinkingAndroid"
    ],
    "hash": "81744a715b858aba3f3f7191487b1bec19a00bb0"
  }
]

---

Generated by PR labeler 🤖

[Kudo]

nix: it looks like the ownership of ThreadSafeJNIGlobalRef is actually not be shared. maybe we could consider unique_ptr a little bit.

[lukmccall]

nix: it looks like the ownership of ThreadSafeJNIGlobalRef is actually not be shared. maybe we could consider unique_ptr a little bit.

When capturing a unique_ptr in a std::function, the generated object loses the copy constructor, making it impossible to convert the lambda to std::function.