Remove InstructionSet_VectorT256 if AVX2 is disabled via env var

[EgorBo]

Fixes #89408

[jkotas]

Is it something that I have somehow regressed by #89342 or was this a pre-existing bug?

[EgorBo]

Is it something that I have somehow regressed by #89342 or was this a pre-existing bug?

Probably, although, the issue reproduces only on CLR (NAOT and R2R are fine) and only if the cpu physically supports AVX2 but it's disabled by the config

[tannergooding]

Might be good to have a test checking Vector<T>.IsHardwareAccelerated as well. I'm guessing we're missing that in CpuId.cs and are only checking the size instead

[jkotas]

Probably, although, the issue reproduces only on CLR (NAOT and R2R are fine) and only if the cpu physically supports AVX2 but it's disabled by the config

Thank you for fixing it!

[EgorBo]

Might be good to have a test checking Vector<T>.IsHardwareAccelerated as well. I'm guessing we're missing that in CpuId.cs and are only checking the size instead

IsHardwareAccelerated didn't break, did it? The only way to test it is to use FILECHECK tests and check disasm - I'll think more about it as part of the quality week to improve test coverage for hw acceleration on asm level for SIMD crossplat APIs