Added vnni512; removed forced bmi2 on CPUS with vnni for AVX-512

[maximmasiutin]

Added vnni512; removed forced bmi2 on CPUS with vnni for AVX-512. This will allow testing of NNUE-related code for modern processors.

[noobpwnftw]

Read comments.

[ppigazzini]

The PR has two problems:

1. it will be a regression for workers tested faster with bmi2 than with avx512
2. vnni512 and vnni256 architectures have the same conditions on props

         elif (
            "avx512vnni" in props["flags"]
            and "avx512dq" in props["flags"]
            and "avx512f" in props["flags"]
            and "avx512bw" in props["flags"]
            and "avx512vl" in props["flags"]
            and "x86-64-vnni512" in targets
        ):
            arch = "x86-64-vnni512"
        elif (
            "avx512vnni" in props["flags"]
            and "avx512dq" in props["flags"]
            and "avx512f" in props["flags"]
            and "avx512bw" in props["flags"]
            and "avx512vl" in props["flags"]
            and "x86-64-vnni256" in targets
        ):
            arch = "x86-64-vnni256"