to improve performance on AMD Zen (#2180) applying wjc404's improvement of the DGEMM kernel from #2186

Add vbroadcastsd kernel to dgemm_kernel_4x8_haswell.S

Makefile.arm: remove -march flags

Override special make variables

as seen in https://github.com/xianyi/OpenBLAS/issues/1912#issuecomment-514183900 , any external setting of TARGET_ARCH (which could result from building OpenBLAS as part of a larger project that actually uses this variable) would cause the utest build to fail. 
(Other subtargets appear to be unaffected as they do not use implicit make rules)

MAINT: remove legacy CMake endif()

Replace vpermpd with vpermilpd in the Haswell/Zen zdot microkernel

Update dgemm_kernel_4x8_haswell.S for reducing cache misses

* clean up a case where CMake endif()
contained the conditional used in the
if(), which is no longer needed /
discouraged since our minimum required
CMake version supports the modern syntax

to improve performance on Zen/Zen2 (as demonstrated by wjc404 in #2180)

Update "dgemm_kernel_4x8_haswell.S" for improving performance on zen2 chips

replaced a bunch of vpermpd instructions with vpermilpd and vperm2f128

Change install_name on osx to match linux

Fix surprising behaviour of NO_AFFINITY=0

Fix CMAKE compilation with MinGW32 and add it to Appveyor

(also copy BUNDERSCORE settings for FORTRAN from the corresponding Makefile)

needs to be unset in the cache for the change to have any effect

(and other small fixes to match Makefile behaviour)

Fixes for PGI compiler

power9 cgemm/ctrmm. new sgemm 8x16

Fix build on PPC970 for FreeBSD

FreeBSD needs those macros too.

FreeBSD needs DCBT_ARG=0 as well.

Fix build on FreeBSD/powerpc64.

Signed-off-by: Piotr Kubaj <pkubaj@anongoth.pl>

Fix DTRMMKERNEL register save for power8 64-bit mode (Fix for #2166)