Memory leak when reading co_code attribute of deepfrozen code objects

[kumaraditya303]

Reproducer:

import os
os.path.join.__code__.co_code

Relevant Valgrind Report:

==21416== HEAP SUMMARY:
==21416==     in use at exit: 397 bytes in 1 blocks
==21416==   total heap usage: 19,728 allocs, 19,727 frees, 2,629,532 bytes allocated
==21416== 
==21416== 397 bytes in 1 blocks are still reachable in loss record 1 of 1
==21416==    at 0x483B7F3: malloc (in /usr/lib/x86_64-linux-gnu/valgrind/vgpreload_memcheck-amd64-linux.so)
==21416==    by 0x2FA6DD: _PyMem_RawMalloc (obmalloc.c:101)
==21416==    by 0x2FC5CD: PyObject_Malloc (obmalloc.c:713)
==21416==    by 0x287BDC: _PyBytes_FromSize (bytesobject.c:103)
==21416==    by 0x2881A6: PyBytes_FromStringAndSize (bytesobject.c:136)
==21416==    by 0x29F609: _PyCode_GetCode (codeobject.c:1374)
==21416==    by 0x29F69B: code_getcode (codeobject.c:1728)
==21416==    by 0x2A31B2: getset_get (descrobject.c:202)
==21416==    by 0x2F86E6: _PyObject_GenericGetAttrWithDict (object.c:1269)
==21416==    by 0x2F8F14: PyObject_GenericGetAttr (object.c:1357)
==21416==    by 0x2F9864: PyObject_GetAttr (object.c:912)
==21416==    by 0x3B5B16: _PyEval_EvalFrameDefault (ceval.c:3472)
==21416== 
==21416== LEAK SUMMARY:
==21416==    definitely lost: 0 bytes in 0 blocks
==21416==    indirectly lost: 0 bytes in 0 blocks
==21416==      possibly lost: 0 bytes in 0 blocks
==21416==    still reachable: 397 bytes in 1 blocks
==21416==         suppressed: 0 bytes in 0 blocks
==21416== 

This is a 3.12 only regression caused by #93382.

[kumaraditya303]

@Fidget-Spinner Can you cherry-pick my fix #93729 to your backport PR #93493?

[Fidget-Spinner]

Thanks for catching this, and sure thing.