feat: Lock Array and Object wrappers for safe multithreading

[mrousavy]

Adds locks to array and object wrappers to make multi-threaded access (read or write operations, e.g. push() and pop()) safe.

Uses a recursive_mutex, might have a small performance impact.

[chrfalch]

I finally got around to read the documentation about concurrency and containers in the C++ standard, and I think we could avoid a lot of the locking in the PR - read more here - I'll try to go through and "mark" methods that we need to be careful about a bit later:

https://en.cppreference.com/w/cpp/container#Thread_safety

[mrousavy]

@chrfalch which methods are you talking about? Consider the following code:

const array = useSharedValue([35, 75234, 385, 357])

const threadA = useWorklet(context1, () => {
  'worklet'
  for (let i = 0; i < array.value.length; i++) {
    console.log(array.value[i])
  }
})

const threadB = useWorklet(context2, () => {
  'worklet'
  while (array.value.length > 0) {
    const item = array.value.pop()
    console.log(item)
  }
})

threadA()
threadB()

In threadB we remove all items, and in threadA we iterate through the items with said index.

Without throwing a lock on pop and index, we would probably encounter issues here.

[mrousavy]

Also, check out how pop() is implemented:

react-native-worklets-core/cpp/wrappers/WKTJsiArrayWrapper.h

Lines 85 to 94 in a92ff4f

	JSI_HOST_FUNCTION(pop) {
	// Pop last element from array
	if (_array.empty()) {
	return jsi::Value::undefined();
	}
	auto lastEl = _array.at(_array.size() - 1);
	_array.pop_back();
	notify();
	return lastEl->unwrapAsProxyOrValue(runtime);
	};

We have three separate operations on our std::vector:

1. _array.size()
2. _array.at(..)
3. _array.pop_back()

And for example, after step 1 (size()), another Thread might've already popped an item in the meantime, so the step 2 (at(..)) call might already be out of bounds.

Unless I'm missing something we need to lock on that entire pop() method

[mrousavy]

Without the lock PR, I get this error randomly after leaving the app run for a while where arrays are accessed from multiple threads: