CpuMath Enhancement: Make bound checking of loops in hardware intrinsics more efficient

[briancylui]

Style changes needed to solve part of #823

Details

• In src\Microsoft.ML.CpuMath\SseIntrinsics.cs and src\Microsoft.ML.CpuMath\AvxIntrinsics.cs, changing while (pDstCurrent + 4 <= pDstEnd) for the loop bound checking into while (pDstCurrent <= pDstEnd - 4) to save an instruction (ref: Port all active C# hardware intrinsics APIs for SSE from SIMD native algorithms #668 (comment))
• It may probably be a CoreCLR issue

[danmoseley]

As well as doing this, we should make sure there's a code gen issue open.

[rosebyte]

Could you provide evidence how much performance we get by the optimisation?

[dotnet-policy-service]

This issue has been marked needs-author-action and may be missing some important information.