internal: schedule buffer to release later instead of leaking

[Person-93]

If a buffer is dropped during gc traversal, use Py_AddPendingCall to release it later instead of leaking it.

Hopefully this can also unblock #6261.

[codspeed-hq]

Merging this PR will improve performance by 17.61%

⚠️ Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

⚡ 3 improved benchmarks
✅ 138 untouched benchmarks
⏩ 6 skipped benchmarks¹

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	Simulation	extract_i128_small_pos	2.6 µs	2.2 µs	+18.16%
⚡	Simulation	extract_i128_zero	2.6 µs	2.2 µs	+17.63%
⚡	Simulation	extract_i128_small_neg	2.7 µs	2.3 µs	+17.04%

Tip

Curious why this is faster? Comment @codspeedbot explain why this is faster on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing Person-93:fix-leak-pybuffer (b409b76) with main (34f08c3)}

Footnotes

1. 6 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[davidhewitt]

Crafty solution! I'm a little wary about Py_AddPendingCall being safe to call from within GC traverse, I think it should be possible to write a test so we can verify.

I would also be quite happy with getting #6330 done and ultimately declaring GC traversal functions unsafe to implement and completely throw away any check from this Drop impl.

[Person-93]

I'm a little wary about Py_AddPendingCall being safe to call from within GC traverse

I think you're right about this.