gh-117139: Fix a few wrong steals in bytecodes.c

[Fidget-Spinner]

I actually caught these using a tool I built following the stackref semantics here https://github.com/faster-cpython/ideas/blob/kenjin/stackref-semantics/3.14/stackref_semantics.md.

I'll send the PR for the tool in a separate PR.

These bugs were quite hard to notice, I will explain each one in comments below.

• Issue: Set up tagged pointers in the evaluation stack #117139

[Fidget-Spinner]

PyObject_SetItem and the entire family actually takes borrowed references to the container and item. So this steal is wrong, it should be borrow.

This was caught when the PyStackRef_CLOSE(v) below tried to close an already invalid stackref (it was made invalid by the steal).

[Fidget-Spinner]

Same for PySet_Add

[Fidget-Spinner]

PyObject_SetAttr actually takes borrowed refs too, so this shouldn't be stolen. This was caught when the DECREF_INPUTS tried to close an already stolen stackref.

[bedevere-bot]

⚠️⚠️⚠️ Buildbot failure ⚠️⚠️⚠️

Hi! The buildbot AMD64 Debian root 3.x has failed when building commit e6543da.

What do you need to do:

1. Don't panic.
2. Check the buildbot page in the devguide if you don't know what the buildbots are or how they work.
3. Go to the page of the buildbot that failed (https://buildbot.python.org/all/#builders/345/builds/8384) and take a look at the build logs.
4. Check if the failure is related to this commit (e6543da) or if it is a false positive.
5. If the failure is related to this commit, please, reflect that on the issue and make a new Pull Request with a fix.

You can take a look at the buildbot page here:

https://buildbot.python.org/all/#builders/345/builds/8384

Failed tests:

• test_repl

Failed subtests:

• test_asyncio_repl_is_ok - test.test_repl.TestInteractiveInterpreter.test_asyncio_repl_is_ok

Summary of the results of the build (if available):

==

Click to see traceback logs

Traceback (most recent call last):
xiting asyncio REPL...
  File "/root/buildarea/3.x.angelico-debian-amd64/build/Lib/asyncio/__main__.py", line 109, in run
    raise OSError(errno.ENOTTY, "tty required", "stdin")
Fatal Python error: _enter_buffered_busy: could not acquire lock for <_io.BufferedWriter name='<stderr>'> at interpreter shutdown, possibly due to daemon threads
Python runtime state: finalizing (tstate=0x000055da69db5ad0)


Traceback (most recent call last):
  File "/root/buildarea/3.x.angelico-debian-amd64/build/Lib/test/test_repl.py", line 199, in test_asyncio_repl_is_ok
    assert_python_ok("-m", "asyncio")
    ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
  File "/root/buildarea/3.x.angelico-debian-amd64/build/Lib/test/support/script_helper.py", line 180, in assert_python_ok
    return _assert_python(True, *args, **env_vars)
  File "/root/buildarea/3.x.angelico-debian-amd64/build/Lib/test/support/script_helper.py", line 165, in _assert_python
    res.fail(cmd_line)
    ~~~~~~~~^^^^^^^^^^
  File "/root/buildarea/3.x.angelico-debian-amd64/build/Lib/test/support/script_helper.py", line 75, in fail
    raise AssertionError("Process return code is %d\n"
    ...<13 lines>...
                            err))
AssertionError: Process return code is -6
command line: ['/root/buildarea/3.x.angelico-debian-amd64/build/python', '-X', 'faulthandler', '-I', '-m', 'asyncio']


Traceback (most recent call last):
xiting asyncio REPL...
  File "/root/buildarea/3.x.angelico-debian-amd64/build/Lib/asyncio/__main__.py", line 109, in run
    raise OSError(errno.ENOTTY, "tty required", "stdin")
Fatal Python error: _enter_buffered_busy: could not acquire lock for <_io.BufferedWriter name='<stderr>'> at interpreter shutdown, possibly due to daemon threads
Python runtime state: finalizing (tstate=0x000055bec2fccad0)

[Fidget-Spinner]

It looks unrelated and more like a buildbot problem.