check for void return FFI functions, fix incorrect cases

[davidhewitt]

This PR attempts to check whether pyo3-ffi and the C headers agree on functions having no return value. The more general case of type checking is hard (pyo3-ffi and bindgen may have different logical types; what we really need to check is ABI compatibility), but checking for void return turns out to be easy enough.

Subsequently I fix two cases which turned out to be wrong in pyo3-ffi.

[davidhewitt]

Need no changelog entry for this one as it'll be newly released in 0.29.

[codspeed-hq]

Merging this PR will improve performance by 27.92%

⚠️ 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

⚡ 29 improved benchmarks
✅ 96 untouched benchmarks

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
⚡	test_class_attribute	16 µs	12.6 µs	+27.44%
⚡	test_class_method	18.4 µs	15.6 µs	+18.62%
⚡	test_empty_class_init	24.7 µs	15.8 µs	+56.09%
⚡	test_empty_class_init_py	14.4 µs	10.8 µs	+33.18%
⚡	test_getter	18.3 µs	15.2 µs	+20.45%
⚡	test_method_call	17.3 µs	13.9 µs	+24.35%
⚡	test_method_call_py	14.9 µs	11 µs	+35.88%
⚡	test_proto_call	14 µs	10.4 µs	+35.15%
⚡	test_proto_call_py	15.3 µs	11.6 µs	+31.74%
⚡	test_setter	18.1 µs	14.7 µs	+22.84%
⚡	test_static_method	17.6 µs	14.6 µs	+20.61%
⚡	test_args_kwargs_py	18.6 µs	15.1 µs	+23.65%
⚡	test_args_kwargs_rs	14.8 µs	11.2 µs	+32.11%
⚡	test_many_keyword_arguments_py	19.3 µs	16.2 µs	+19.15%
⚡	test_many_keyword_arguments_rs	20.2 µs	16.7 µs	+20.91%
⚡	test_none_py	14.4 µs	10.3 µs	+39.16%
⚡	test_none_rs	13.4 µs	9.7 µs	+38.36%
⚡	test_positional_only_py	13.9 µs	10.3 µs	+34.6%
⚡	test_positional_only_rs	14.4 µs	10.7 µs	+34.45%
⚡	test_simple_args_kwargs_py	20.5 µs	17.2 µs	+19.27%
...	...	...	...	...

ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.

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 davidhewitt:void-return-ffi (6c136d0) with main (c92117c)}