[Bug] JVM crash when call a method of a PyObject in python async thread, which returns a custom java object.

[wenjin272]

1. In pemja python thread, we get a java object, and then we passed it to a python asynchronous thread.
2. In the async thread, when we call the method of this java object, if the method returns a custom java object, the jvm will crash; If the method returns jdk built-in object, like string or List, it will work normally.

How to reproduce

clone https://github.com/wenjin272/flink-agents/tree/built-in-async
./tools/build.sh
pip install apache-flink==1.20.3
uv run pytest python/flink_agents/e2e_tests/e2e_tests_integration/react_agent_test.py::test_react_agent_on_remote_runner

Here is the err file:
long_text_a1b36098-7418-4bdb-b9cf-ed8e7469ecb2.txt

[bgeng777]

I believe the root cause is that when using custom java object in python threads which are not managed by pemja (e.g. spawned by python's asyncio), the necessary jcpThread is not initialized correctly by calling JcpPy_InitThread().
Some internal types can work because they are processed by methods like JcpPyDict_FromJMap.

I created a PR for this. I verified it via a local test
@wenjin272 Could you please give it a try for this fix?
I met errors like [ERROR] 符号: 类 DataStream [ERROR] 位置: 类 org.apache.flink.agents.api.AgentsExecutionEnvironment when I tried to run ./tools/build.sh command. It looks like a environment problem of my mac.