Update BitArray.CopyTo to use cross-platform intrinsics

[Copilot]

Fixes #114818 (comment)

This PR updates BitArray.CopyTo to use cross-platform intrinsic APIs instead of platform-specific ones, making the code more portable across different hardware architectures.

Changes

AVX512 path (Vector512)

• Avx512BW.IsSupported → Vector512.IsHardwareAccelerated
• Avx512BW.Shuffle(x, y) → Vector512.Shuffle(x, y)
• Avx512F.And(x, y) → x & y
• Avx512BW.Min(x, y) → Vector512.Min(x, y)
• Avx512F.Store(x, y) → y.Store(x)

AVX2 path (Vector256)

• Avx2.IsSupported → Vector256.IsHardwareAccelerated
• Avx2.Shuffle(x, y) → Vector256.Shuffle(x, y)
• Avx2.And(x, y) → x & y
• Avx2.Min(x, y) → Vector256.Min(x, y)
• Avx.Store(x, y) → y.Store(x)

Unified Vector128 path

• Consolidated SSSE3, PackedSimd, and AdvSimd paths into a single unified Vector128 implementation
• Ssse3.IsSupported → Vector128.IsHardwareAccelerated
• Ssse3.Shuffle(x, y) → Vector128.Shuffle(x, y) (internally uses PackedSimd.Swizzle on WASM and AdvSimd.Arm64.VectorTableLookup on ARM)
• Sse2.And(x, y) / PackedSimd.And(x, y) / AdvSimd.And(x, y) → x & y
• Sse2.Min(x, y) / PackedSimd.Min(x, y) / AdvSimd.Min(x, y) → Vector128.Min(x, y)
• Sse2.Store(x, y) / PackedSimd.Store(x, y) / AdvSimd.Arm64.StorePair(x, y, z) → y.Store(x)
• Removed ~63 lines of redundant platform-specific code

Benefits

• Cross-platform compatibility: These cross-platform intrinsics provide the same functionality as the platform-specific ones but allow the JIT to optimize for the actual hardware at runtime
• Improved maintainability: Single Vector128 implementation instead of three separate paths (SSSE3, PackedSimd, AdvSimd)
• Better ARM performance: Using VectorTableLookup instead of manually chained ZipLow/ZipHigh instructions
• Reduced code duplication: Consolidated multiple platform-specific paths into unified implementations

Testing

All existing System.Collections tests pass (33,438 tests, 0 failures).

Fixes #116079

Original prompt

---

This section details on the original issue you should resolve

<issue_title>Update BitArray.CopyTo to use the xplat intrinsics</issue_title>
<issue_description>As per #114818 (comment), BitArray.CopyTo could benefit from using the cross platform intrinsic APIs rather than the platform specific ones.

This can be achieved by swapping out these API calls:

• Avx512BW.IsSupported -> Vector512.IsHardwareAccelerated
• Avx512BW.Shuffle(x, y) -> Vector512.Shuffle(x, y)
• Avx512F.And(x, y) -> x & y
• Avx512BW.Min(x, y) -> Vector512.Min(x, y)
• Avx512F.Store(x, y) -> y.Store(x)

Similar changes could also be made to the Avx2 path (using Vector256) and so on.</issue_description>

Comments on the Issue (you are @copilot in this section)

Fixes #116079

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.

[dotnet-policy-service]

Tagging subscribers to this area: @dotnet/area-system-collections
See info in area-owners.md if you want to be subscribed.

[Copilot]

Pull Request Overview

This PR updates BitArray.CopyTo to use cross-platform intrinsic APIs instead of platform-specific ones, improving code portability across different hardware architectures. The changes replace platform-specific AVX512 and AVX2 intrinsics with their cross-platform equivalents.

Key changes:

• Replace AVX512BW/AVX512F intrinsics with Vector512 cross-platform APIs
• Replace AVX2/AVX intrinsics with Vector256 cross-platform APIs
• Use hardware acceleration detection and operator overloads for better portability

[tannergooding]

It's likely worth noting that since this is byte, there shouldn't be any considerations. However, Vector.Shuffle has to have deterministic behavior across platforms for "out of range" indices and so, for other T, can slightly pessimize to handle that scenario.

We're using a "vector constant" here, but its indirect and through a couple local copies. The JIT should still see it in such other scenarios and the cost is minimal, so shouldn't matter. But it's something to note anyways.