Make FFI Pointers a bit more typesafe

[Person-93]

Objective

Simplify casting back and forth when unsafely converting raw pointers to pyo3 smart pointers. It's a bit error prone and makes the code appear very cluttered.

// this
let args: Borrowed<'py, 'py, PyTuple> =
     unsafe { Borrowed::from_ptr(py, args).cast_unchecked::<PyTuple>() };
let kwargs: Option<Borrowed<'py, 'py, PyDict>> =
    unsafe { Borrowed::from_ptr_or_opt(py, kwargs).map(|kwargs| kwargs.cast_unchecked()) };

// becomes this
let args = unsafe { args.cast::<FfiTypeOf<PyTuple>>().assume_borrowed(py) };
let kwargs = unsafe { kwargs.cast::<FfiTypeOf<PyDict>>().assume_borrowed_or_opt(py)};

The public API of smart pointers is expanded as well. They can be unsafely created from any pointer which implements FfiObj and the T param is the associated `ApiObj.

Changes

• Add two hidden traits ApiObj and FfiObj
• Update macros for implementing native type info to optionally implement those traits
• Update pyclass macro to implement those traits
• implement existing FfiPtrExt trait for all *mut T where T: FfiObj
• Remove type casts from a few places in the codebase

[Person-93]

I'm having trouble building the test_utils mod. https://github.com/PyO3/pyo3/actions/runs/35670200524/job/106564800184?pr=6445#step:7:1253

[codspeed]

Merging this PR will degrade performance by 11.18%

⚠️ 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

❌ 2 regressed benchmarks
✅ 141 untouched benchmarks
⏩ 6 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
❌	into_i128_zero	817.2 ns	926.9 ns	-11.84%
❌	into_i128_small_pos	932.8 ns	1,042.5 ns	-10.52%

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing Person-93:ffi-ptr-more-typesafe (25864be) with main (1655cdf)}

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. ↩

[mejrs]

These two feel ~similar in terms of how error prone and cluttered they feel. I'm not sure it's worth the churn.