Optimize list materialization across element types

[skuirrels]

What changed

• add cached typed materializers for previously boxed List<T> fallback paths
• preserve the established specialized fast paths for int, float, double, and decimal
• retain natural object materialization for nested/reference element types
• preserve converted enum semantics for enum and nullable-enum LIST values
• add regression coverage for pre-sized List<long> and nullable long values
• add deterministic nullable-enum LIST regression coverage
• add a 16-case BenchmarkDotNet matrix covering nullable numerics, nested lists, and fixed arrays

Why

Generic LIST and ARRAY materialization previously fell through IList.Add(object?), boxing value-type elements and increasing allocations. The typed strategy removes that per-element boxing without routing established specialized types through an extra interface dispatch.

Evidence

• targeted reader/type tests: 98/98 on .NET 8 and .NET 10
• full test suite: 7,024/7,024 on .NET 8 and .NET 10
• Release solution build succeeded
• exact 1.5.5.3 fork packages validated and consumer-smoke-tested on .NET 8 and .NET 10
• 16-case before/after benchmark matrix found no observed regressions
• List<long>: 17.80 ms to 13.97 ms; 57.23 MB to 20.60 MB
• List<long?>: 21.39 ms to 17.87 ms; 63.33 MB to 35.86 MB
• List<DateTime>: 41.02 ms to 32.22 ms; 57.23 MB to 20.60 MB
• List<Guid>: 28.94 ms to 23.85 ms; 81.64 MB to 32.81 MB

Release

Intended for stable fork release v1.5.5.3 after hosted checks pass.