Move some x86_64 DYNAMIC_ARCH targets to new DYNAMIC_OLDER option

Fixed a few more unnecessary calls to num_cpu_avail.

Create OpenBLASConfig.cmake in cmake builds as well

Revert "Use usleep instead of sched_yield by default"

* Return a somewhat sane default value for L2 cache size if cpuid returned something unexpected

Fixes #1610, the KVM hypervisor on Google Chromebooks returning zero for CPUID  0x80000006, causing DYNAMIC_ARCH
builds of OpenBLAS to hang

I don't have as many benchmarks for these as for gemm, but it should still
make a difference for small matrices.

Enable parallel make on MS Windows by default

fixes #874

Improve performance of GEMM for small matrices when SMP is defined.

Always checking num_cpu_avail() regardless of whether threading will actually
be used adds noticeable overhead for small matrices.  Most other uses of
num_cpu_avail() do so only if threading will be used, so do the same here.

Use a single thread for small input size in zaxpy

Use usleep instead of sched_yield by default

copies daxpy improvement from #27, see #1560

sched_yield only burns cpu cycles, fixes #900,  see also #923, #1560

Initial support for SkylakeX / AVX512

Improved AVX512 test case for c_check

clang 3.4 managed to accept the original test code, only to fail on the actual Skylake asm later

Restore _Atomic define before stdatomic.h for old gcc

see #1593

Check build system support for AVX512 instructions

Fix inverted condition in _Atomic declaration

fixes #1593

Disable AVX512 (Skylake X) support if the build system is too old

this required switching to the generic gemm_beta code (which is faster anyway on SKX)
for both DGEMM and SGEMM

Performance for the not-retuned version is in the 30% range