Add batch primitives and generated accessors

[RejectKid]

Summary

• add checked, allocation-free BitBatchPrimitives APIs for packed and record-strided reads and writes
• cover all supported signed/unsigned integral types, native integers, Booleans, and LSB/MSB order with 88 symmetric overloads
• validate the complete batch before processing and use bounded tail handling so exact-size buffers need no unsafe padding
• add opt-in BitObjectAttribute.GenerateBatchAccessors generation for strongly typed integral, Boolean, and enum helpers
• generate packed and strided static readers plus writers for non-readonly fields while preserving declared accessibility
• keep generated batch helpers checked even when instance accessors opt into unsafe mode
• document direct and generated usage, validation rules, and the maintained-fork feature overview
• add functional, generator, boundary, API-surface, and performance coverage

Example

[BitObject(BitOrder.LeastSignificant, GenerateBatchAccessors = true)]
public partial struct Packet
{
    [BitField("Id", 12, BitFieldType.UInt16)]
    [BooleanField("Enabled")]
    private ushort _layout;
}

Packet.ReadIdBatch(records, bitStride: 16, destination: ids);
Packet.ReadEnabledBatch(records, bitStride: 16, destination: enabled);
Packet.WriteIdBatch(records, bitStride: 16, values: ids);

Performance

AMD Ryzen 9 9900X, .NET 10.0.8, BenchmarkDotNet MediumRun with 15 measured iterations, 10 warmups, and 2 launches. Values are normalized per field; all cases allocate 0 B.

12-bit UInt32 operation	Scalar loop	Direct batch	Generated batch	Result
Packed read	0.8177 ns	0.7882 ns	0.7788 ns	direct 3.6% faster; generated 4.8% faster
32-bit-strided read	0.8586 ns	0.7697 ns	—	10.4% faster
Packed write	2.7427 ns	2.7273 ns	2.7368 ns	effectively tied

The generated wrapper is fully inlined and adds no measurable overhead. Write batching is included for API symmetry and ergonomics; the measured difference is within noise.

Validation

• Release solution build: 0 warnings, 0 errors
• .NET 8 tests: 4,602 passed
• .NET 10 tests: 4,602 passed
• randomized/reference-equivalence coverage for every integral reader, both bit orders, packed/strided layouts, signed values, Booleans, enums, exact-size tails, and invalid ranges
• generated packed/strided reader and writer coverage, including readonly behavior
• public API reflection audit: 44 read and 44 write overloads
• focused BenchmarkDotNet MediumRun: passed, 0 B allocated
• NuGet pack and validation using 1.6.0-preview.2: passed
• targeted formatting verification and git diff --check: passed