As observed using GCC 10 using -march=native -ftree-vectorize
on Knights Landing, it is now smart enough to find clobbers inside
non-inlined static functions.

In particular, sgemv counted on a kernel to preserve the whole
%ymm2 register (since it was not in the clobber list), but the top
part was destroyed by vzeroupper. This caused many tests to fail.

This patch makes sure all xmm (and ymm/zmm by extension) registers
are listed as clobbered to avoid this happening, as most kernels
already did correctly in fact.

Clean up duplicate definitions in POWER8 kernels and fix power10 option passing

The
`#if defined(SKYLAKEX) || defined (COOPERLAKE)`
from that commit was before #include "common.h" so caused the
compiled function to be empty, returning garbage results for
qualifying sgemm's on those architectures.

Closes #2914

Add compiler flag for SSE4 where available

Revert special handling of Windows xNRM2 and enable C+intrinsics kern…

As the new MMA instructions need the inputs in 4x2 order for bfloat16,
changing the format in copy/packing code.  This avoids permute instructions
in the gemm kernel inner loop.

This patch fixes compilation errors due to recent renaming from SH to SB
with BUILD_BFLOAT16.

Rename "HALF" precision functions (sh prefix) to "BFLOAT16" with "sb" prefix

add fninit to reset fpu registers before assembler routines

Use generic C for (D/Z)NRM2 on Windows x86_64

Support building only a subset of variable types

Optimize the performance of dot by using universal intrinsics in X86/ARM