After a fork it is possible that blas_thread_buffer has already
allocated memory buffers: goto_set_num_threads does allocate those
already and it may be called by num_cpu_avail in case the OpenBLAS
NUM_THREADS differ from the OMP num threads.
This leads to a memory leak which can cause subsequent execution of BLAS
kernels to fail.

Fixes #2993

Fix sum optimize issues

Optimize the performance of rot by using universal intrinsics

Fix missing -mfma compiler flag in cmake builds without DYNAMIC_ARCH

rebase

Improve the performance of rot by using AVX512 and AVX2 intrinsic

Fix missing AVX2 and FMA3 capabilities in FORCE_target mode

Optimize sdot/ddot for POWER10

Fix handling of cpu capability flags in DYNAMIC_ARCH builds

This patch makes use of new POWER10 vector pair instructions for
loads and stores.

rebase

Fix macro name used in ifdef for POWERPC/PGI

Ensure that EXPRECISION is disabled for DYNAMIC_ARCH with TARGET=GENERIC and fix CMAKE DYNAMIC_ARCH builds

POWER10:  Change dgemm unroll factors

NO_EXPRECISION is disabled for the GENERIC_TARGET already, so prevent mixing with code parts that use a different float size by default

add openbsd 68+ gfortran name

Reunify default BUFFER_SIZE on ARM64 to avoid crashes in DYNAMIC_ARCH mode

Changing the unroll factors for dgemm to 8 shows improved performance with
POWER10 MMA feature.   Also made some minor changes in sgemm for edge cases.

rebase

fix avx2 detection

reword commits to make it clearer