[bug] Fix crash when using custom java object in python async thread

[bgeng777]

For #93 . When a Java object is passed from a pemja thread to a Python async thread, calling a
method that returns a custom Java object crashes the JVM with SIGSEGV. Methods
returning built-in types (String, List, etc.) work normally.

# SIGSEGV (0xb) — JcpPyJObject_New+0x158
# NULL dereference at 0x40 (offset of name_to_attrs in JcpThread)

Cause

pyjobject_init (pyjobject.c) calls JcpThread_Get() to access the per-thread
attribute cache. On non-pemja threads, this returns NULL — which is immediately
dereferenced without a NULL check:

jcpThread = JcpThread_Get();           // NULL on async threads
if (jcpThread->name_to_attrs == NULL)  // SIGSEGV

Built-in types bypass this path — they have dedicated converters in
JcpPyObject_FromJObject that never call JcpThread_Get(). Only custom objects,
Collection, Iterable, and Iterator go through JcpPyJObject_New → pyjobject_init.

Fix

File: src/main/c/pemja/core/python_class/pyjobject.c

When JcpThread_Get() returns NULL, clear the Python error and skip the per-thread
cache — build the attrs dict directly and manage reference counts without the cache.

• pemja threads: unchanged, cache works as before.
• Non-pemja threads: custom Java objects now work, with a minor cost of repeated
  JNI reflection (no cache).

[wenjin272]

pemja/src/main/c/pemja/core/pylib.c

Line 191 in 2c84a2f

if (!ret && !PyErr_Occurred()) {

I tested this patch locally and found that an exception is thrown first when JcpThread is NULL. After commenting out this logic, it appears to work.