Use the inline definition of Py_TYPE in the unlimited API on 3.14+

[exg]

No description provided.

[codspeed-hq]

Merging this PR will improve performance by 22.31%

⚡ 10 improved benchmarks
✅ 130 untouched benchmarks

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
⚡	extract_u128_small	1,200.8 ns	956.9 ns	+25.49%
⚡	extract_u128_zero	1,230 ns	986.1 ns	+24.73%
⚡	extract_i128_small_pos	1.3 µs	1 µs	+24.02%
⚡	extract_i128_zero	1.3 µs	1 µs	+23.35%
⚡	extract_u128_u32_max	1.3 µs	1.1 µs	+23.03%
⚡	extract_i128_small_neg	1.4 µs	1.1 µs	+22.02%
⚡	extract_u128_u64_max	1.4 µs	1.2 µs	+20.83%
⚡	extract_u128_max	1.4 µs	1.2 µs	+20.29%
⚡	extract_i128_pos_max	1.5 µs	1.2 µs	+19.79%
⚡	extract_i128_neg_min	1.5 µs	1.2 µs	+19.76%

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 exg:inline-py-type-in-unlimited-api (cfe0f23) with main (e2f8fb9)}

[ngoldbaum]

@exg mentioned on discord that they saw ~10% speedup in an ormsgpack benchmark with this change.

@davidhewitt does using the direct field access on version-specific builds seem OK too you?

[davidhewitt]

This makes sense to me, and is consistent with the headers.

(Seems like exactly the sort of performance advantage permissible from the version-specific builds.)